What AI chatbots say they do with a lab report you upload
People now paste lab reports into AI chatbots as a matter of routine, and a growing number of apps exist only to read an uploaded report. We wanted to know what each company says, in its own published words, about what happens to that report afterwards. So we read the privacy policies, terms, and help pages of 7 chatbots and 12 lab-result apps and wrote down the answer to the same questions for each, with the sentence it came from.
This page records what the companies say, as of November 10, 2026. We did not test what any of them does. Every answer below has a quotation and a link, and where a company’s pages say nothing on a question, the answer is “no statement found” and not “no.” How we did this and what it cannot tell you are at the end.
The short version
- None of the 7 chatbots says HIPAA protects a report you upload. Where a company addresses the law at all, it says the law does not apply. Perplexity writes of its own health feature: “Perplexity Health is designed for personal wellness. HIPAA does not apply to consumer health products like Perplexity Health.”
- Most chatbots say your chats can train their models unless you change a setting. One, Meta AI, has no opt-out for US users that we could find.
- Deleting a chat does not always delete the report. ChatGPT keeps uploaded files in a separate Library, Gemini keeps chats a reviewer has seen for up to three years, and Meta’s terms say deleting “may not delete our copy.”
- Five of the 7 chatbots say staff or contractors may read what you send, and five tell you in writing not to send anything sensitive. Anthropic says its employees cannot open conversations by default, with two exceptions. Perplexity’s pages do not say.
- 7 of the 12 lab-result apps advertise HIPAA compliance. Only 2 of those say whether the law applies to them at all, and those say it does not or may not.
- 3 apps carry an App Store label reading “Data Not Collected” while their own privacy policies list the data they collect.
Why HIPAA is the wrong thing to look for
HIPAA is the US law that protects your medical records while a doctor, hospital, lab, or health plan holds them, and while a company working for one of those holds them under contract. It does not follow a copy you upload yourself. A chatbot or an app you sign up for on your own is, in almost every case, outside it. We covered the background in how to use AI on your lab results without uploading them.
That makes “HIPAA compliant” on a consumer app’s home page hard to interpret. The phrase can describe real security practices. It does not mean you have the rights HIPAA gives a patient, and it does not mean a regulator can hold the company to that law for your upload. What governs your report is the company’s privacy policy, which is why we read those.
The chatbots
| Chatbot | Used to train AI? | How long is it kept? | Can people read it? | Used for ads or sold? | What it says about HIPAA | Health data handled differently? |
|---|---|---|---|---|---|---|
| ChatGPT | Used for training unless you opt out | Kept until deleted; uploaded files saved separately | Yes: staff and contractors, for listed reasons | Ads on Free and Go; data not sold | No statement found | Separate Health feature; ordinary chats follow standard rules |
| Claude | Used unless you opt out | Until you delete; then 30 days | Not by default; two exceptions | No ads; data not sold | Individual plans cannot enable HIPAA | Separate health data policy; scope unclear |
| Copilot | Not used to train foundation models | Chats kept; no time limit given | Yes; terms warn of human review | Used for ads without Microsoft 365 | No statement found | State health-data notice covers Copilot prompts |
| Gemini | On by default; can be turned off | 18 months default; reviewed chats 3 years | Yes; reviewers include contractors | Not used for ads; not sold | No statement found | No health mode; general rules only |
| Grok | May be used; opt-out when logged in | Until you delete; then 30 days | Yes; authorized staff may review chats | Not sold; ad use not addressed | No statement found | Asks users not to share health data |
| Meta AI | Used to improve AI; no US opt-out found | No fixed period; deletion may leave Meta's copy | Yes: human review, including outside vendors | AI conversations personalize ads; not sold | No statement found | No health mode; warns against sharing sensitive topics |
| Perplexity | On by default; can be turned off | Files 30 days; chats while account active | No statement found | Not sold for money; some data to advertisers | No statement found | Health data notice; separate Perplexity Health feature |
Each row covers the ordinary consumer product on a personal account in the United States. Business, school, and developer versions run under different terms, often stricter ones. Open a name below for the full answer, the quotation behind it, and the limits of what we could check.
ChatGPT (OpenAI)
What we looked at: Free, Go, Plus and Pro personal plans, signed in, United States
Used to train AI? Used for training unless you opt out. OpenAI's help pages say content on personal plans, including uploaded files, may be used to train its models unless you turn off "Improve the model for everyone" in Settings. Opting out covers new conversations only. Temporary chats are not used.
“OpenAI may use content (including files you upload) submitted to ChatGPT and other services for individuals to improve model performance if you have “Improve the model for everyone” turned on. You can turn it off anytime in Settings > Data Controls.”
Using Library to manage files in ChatGPT; saved from a browser as a PDF
How long is it kept? Kept until deleted; uploaded files saved separately. The help pages say chats stay until you delete them, then are scheduled for permanent deletion within 30 days. Uploaded files are also saved automatically to your Library; deleting the chat does not delete that copy, which must be deleted separately.
“Deleting a chat containing files does not delete a separate file that remains saved in Library. To remove a saved Library file, delete it from Library.”
Using Library to manage files in ChatGPT; saved from a browser as a PDF
Can people read it? Yes: staff and contractors, for listed reasons. The help page says authorized OpenAI staff and outside service providers may access your content, which includes files, to investigate abuse, give support, handle legal matters, or improve models. Only the last reason is switched off by opting out of training.
“A limited number of authorized OpenAI personnel, as well as trusted service providers that are subject to confidentiality and security obligations, may access user content only as needed for these reasons:”
How OpenAI handles data in consumer services; saved from a browser as a PDF
Used for ads or sold? Ads on Free and Go; data not sold. The ads page says Free and Go plans may show ads based on your current chat; Plus and Pro have none. It says advertisers never receive chats and ads are not eligible to appear near personal health topics. The policy says data is not sold.
“If you turn off ads personalization, you will still see ads based on the context of your current chat thread, but we won't use other chat threads, ads history or topics to inform the ads we show you.”
Ads in ChatGPT; saved from a browser as a PDF
What it says about HIPAA No statement found. HIPAA is not mentioned in the US privacy policy or the help pages we saved about ordinary ChatGPT. The only statement is about the separate Health feature, which its help page says is not intended for covered-entity use.
Health data handled differently? Separate Health feature; ordinary chats follow standard rules. A US-only Health feature has its own privacy notice. Its help page says connected medical records, and chats that use them, are not used for training or ad targeting. A lab report uploaded to an ordinary chat follows your standard settings instead.
“No. Connected medical records and Apple Health information, and conversations that use it, are not used to train our foundation models or target ads, regardless of the model-training setting you choose for ChatGPT.”
Health in ChatGPT; saved from a browser as a PDF
Also worth knowing
- Saved memories are stored apart from chats, so deleting the chat where you shared a lab result does not delete what ChatGPT remembered from it. “Saved memories are stored separately from chat history. Deleting the original chat does not automatically delete a separate saved memory.” (Memory in ChatGPT)
- OpenAI's own help page tells users not to enter sensitive information they would not want reviewed or used. “Please do not enter sensitive information that you would not want reviewed or used.” (How OpenAI handles data in consumer services)
- OpenAI's only HIPAA statement concerns its Health feature, which it says is not for covered-entity use and comes with no Business Associate Agreement (the contract HIPAA requires of vendors). “Health in ChatGPT is not intended for clinical or covered-entity use and does not offer a Business Associate Agreement.” (Health in ChatGPT)
- The newer Health help page says Health conversations can create ordinary memories, while the June 2026 Health Privacy Notice says the main account will not access memories created within ChatGPT Health. “Yes. When memory is on, conversations that use Health can create memories to personalize future chats.” (Health in ChatGPT)
- Giving a thumbs up or thumbs down can send the whole conversation to model training even after you have opted out. “For individual plans (Free, Plus, and Pro), if you submit feedback on a response—for example, by selecting thumbs up or thumbs down—we may use the entire conversation associated with that feedback to improve our models, even if you have opted out of model training.” (How your data is used to improve model performance)
Limits of what we checked
- OpenAI's sites blocked automated access, so every quote comes from PDFs of the live pages saved from a browser on 2026-11-09 or 2026-11-10 and matched against the extracted text.
- The saved PDFs do not print their web addresses. Source links are the likely addresses of the same-titled pages and were not confirmed from the PDFs, except the Memory page, whose address was read from a link inside the PDF.
- Help-centre pages show only relative dates (such as "Updated: 21 days ago"), so their page dates are left blank. Approximate dates: Library article 2026-10-27, consumer-data and training articles 2026-10-19, ads article 2026-11-02, Health article 2026-10-26, Memory article 2026-10-31.
- No saved page uses the word "default" for the training setting on personal plans or states the switch's starting position, and the training article names Free, Plus and Pro but not the Go plan.
- The Health Privacy Notice (June 29, 2026) and the Health help page (about October 2026) describe different versions of the Health feature; where they differ we describe the newer help page.
- Not stated in the saved pages: whether the content of an uploaded file is used to choose ads, whether health-related past chats or memories are kept out of ad personalization, and how long a deleted file stays in "Recently deleted".
- The Memory page says saved memories survive chat deletion; it does not say how quickly information drawn from chat history stops being used after one chat is deleted, or whether deleting a Library file removes what was remembered from it.
- ChatGPT Business, Enterprise, Edu and ChatGPT for Healthcare are sold under other terms and are not covered here.
Claude (Anthropic)
What we looked at: Free, Pro and Max consumer plans on claude.ai, United States
Used to train AI? Used unless you opt out. The privacy policy says chats and uploads may be used to train models unless you opt out in settings. Flagged conversations and feedback you submit are used even if you opt out. The pages do not say which way the switch starts.
“We may use your Inputs and Outputs to train and improve Anthropic AI models, unless you opt out through your account settings.”
Privacy Policy, dated September 10, 2026; downloaded from the company’s site
How long is it kept? Until you delete; then 30 days. The policy says a deleted conversation leaves your history at once and Anthropic's storage within 30 days. A help page says that if you allow training, data may be kept without your identity for up to 5 years.
“You also are able to delete individual conversations, which will be removed immediately from your conversation history and automatically deleted from our back-end within 30 days.”
Privacy Policy, dated September 10, 2026; downloaded from the company’s site
Can people read it? Not by default; two exceptions. A help page says employees cannot open your conversations unless you send feedback or a review is needed to enforce the usage rules, in which case only designated Trust & Safety staff may. Contractors are not mentioned.
“Limited Access: By default, Anthropic employees cannot access your conversations unless: You explicitly consent to share your data with us as a part of giving us feedback (learn more). Review is needed to enforce our Usage Policy.”
How does Anthropic protect the personal data of Claude users?, dated March 16, 2026; downloaded from the company’s site
Used for ads or sold? No ads; data not sold. Anthropic's privacy help page says it does not sell your data and Claude stays ad-free. The privacy policy itself mentions targeted advertising of Anthropic's own products, with an opt-out, and does not say chats are used for it.
“While our products have evolved, our commitments haven't: We don’t sell your data, Claude remains ad-free, and you can control whether your conversations are used to improve Anthropic’s AI models.”
Updates to our Privacy Policy, dated June 8, 2026; downloaded from the company’s site
What it says about HIPAA Individual plans cannot enable HIPAA. The consumer policies do not mention HIPAA, the US health privacy law. A help page says individual plans cannot turn on HIPAA features; the agreement hospitals and clinics need is offered only for Enterprise plans and the developer service.
“You can enable HIPAA from organization settings if your organization is on an Enterprise plan. Team plans and individual plans (Free, Pro, Max, and Ultra) can't enable HIPAA.”
HIPAA-ready Enterprise plans; downloaded from the company’s site
Health data handled differently? Separate health data policy; scope unclear. Anthropic has a Consumer Health Data Privacy Policy for Washington and states with similar laws. Its two descriptions of what the policy covers differ, so it is not clear whether a lab report uploaded into an ordinary chat falls under it.
“This Consumer Health Data Privacy Policy applies only where Anthropic acts as a data controller—for example, when you interact with Claude.ai or other products as a consumer for personal use (“Services”).”
Consumer Health Data Privacy Policy, dated January 12, 2026; downloaded from the company’s site
Also worth knowing
- Opting out of training does not cover chats flagged for safety review or chats you rate with thumbs up or down. “Even if you opt out, we will use Materials for model training when: (1) you provide Feedback to us regarding any Materials, or (2) your Materials are flagged for safety review to improve our ability to detect harmful content, enforce our policies, or advance our safety research.” (Consumer Terms of Service)
- If you allow training, your data may be kept for up to five years with your identity removed. “If you allow us to use your chats or coding sessions to improve Claude, we may retain your data in a de-identified format for up to 5 years in our model training pipelines.” (How long do you store my data?)
- The main privacy policy says the health data policy applies if you connect a health app, while the health policy itself says it also covers health data you provide directly. “If you are located in Washington or a state with similar consumer health data laws, please read our Consumer Health Data Privacy Policy which applies to you if you integrate third party health applications with Claude.” (Privacy Policy)
- Anthropic's promise not to train on health data is made about its health connectors (paid plans, US only), not about a lab report you upload by hand into a normal chat. “These integrations are private by design. Users can choose exactly the information they share with Claude, must explicitly opt-in to enable access, and can disconnect or edit Claude’s permissions at any time. We do not use users’ health data to train models.” (Anthropic news post on healthcare and life sciences)
- The ad-free commitment is made in a blog post and a help page rather than in the privacy policy, and the blog post leaves room to revisit it. “Should we need to revisit this approach, we’ll be transparent about our reasons for doing so.” (Claude is a space to think)
Limits of what we checked
- Enterprise, Team, Education, Government and developer (API) plans run under different terms and are not covered here. The HIPAA quote comes from an Enterprise help article and is used only for what it says about individual plans.
- None of the pages checked says whether the training switch is pre-set to on or off when you are asked to choose at signup.
- No page checked addresses uploaded files separately from chats, for training or for deletion; that an attachment is deleted with its conversation is an inference, not a stated fact.
- The pages checked do not say whether people can read conversations that go into training when you have allowed it, and do not mention contractors as reviewers.
- The HIPAA help page shows no absolute date, so that cell has no page date.
- The health connector pages for Function and Android Health Connect, and the support article on memory, were not opened. An "Ultra" individual plan named on two support pages was not researched.
Copilot (Microsoft)
What we looked at: Updated consumer Copilot app and copilot.microsoft.com with a personal Microsoft account, United States
Used to train AI? Not used to train foundation models. Microsoft's help page says prompts, responses and file contents are not used to train foundation models, the large general-purpose AI models. It does not rule out other uses; the Privacy Statement says conversation data is used to improve Copilot.
“Prompts, responses, and your file contents when using the Microsoft Copilot app aren't used to train foundation models.”
Microsoft Copilot for individuals: your activity history; downloaded from the company’s site
How long is it kept? Chats kept; no time limit given. The new app's help pages say Copilot keeps your past conversations so you can return to them. They give no retention period, no deadline for removing deleted chats and no rule for uploaded files.
“Microsoft Copilot keeps your prior conversations with Copilot so you can revisit them or continue them at any time.”
Microsoft Copilot for individuals: your privacy controls and choices; downloaded from the company’s site
Can people read it? Yes; terms warn of human review. The Copilot Terms of Use say data may be processed by people as well as machines and tell you not to share anything you would not want reviewed. The new help pages do not say who reviews, when, or whether you can opt out.
“Copilot may include both automated and manual (human) processing of data. You shouldn’t share any information with Copilot that you don’t want us to review.”
Copilot Terms of Use, dated August 18, 2026; downloaded from the company’s site
Used for ads or sold? Used for ads without Microsoft 365. Microsoft's pages say people without a Microsoft 365 subscription may see ads, and that ads may relate to the current conversation, including files you share. Personalized ads can be switched off; ads based on the current conversation remain.
“Ads may be shown that relate to the current interaction you are having with us, such as your Copilot conversations (including files you share); your current location; transactions; product usage; search queries; or the content you’re viewing.”
Microsoft Privacy Statement; downloaded from the company’s site
What it says about HIPAA No statement found. We found no mention of HIPAA, the US health privacy law, in the Copilot overview, privacy controls, activity history, Terms of Use or Consumer Health Data Privacy Policy, or in the consumer sections of the Privacy Statement. Only the separate Copilot Health feature mentions it.
Health data handled differently? State health-data notice covers Copilot prompts. Microsoft's Consumer Health Data Privacy Policy, a notice under US state laws, says health-related Copilot prompts are collected. A lab report in an ordinary chat gets no other special handling. Copilot Health is a separate space with stricter rules.
“For example, we collect your Bing search queries and Copilot prompts, which may include queries concerning nutrition, wellness, fitness, medical conditions, or other health-related topics.”
Microsoft Consumer Health Data Privacy Policy; downloaded from the company’s site
Also worth knowing
- Turning off ad personalization does not stop ads chosen from what you are discussing right now. “Generic ads are selected based on the most recent content of your current conversation and don’t use your past chat history.” (Microsoft Copilot for individuals: your privacy controls and choices)
- Microsoft's one stated limit on sensitive data in advertising covers personalized ads only, not ads based on the current conversation. “Sensitive Data: Microsoft Advertising does not collect, process, or disclose personal data that qualifies as “sensitive data” under applicable U.S. state data privacy laws for the purposes of providing personalized advertising.” (Microsoft Privacy Statement)
- The promise not to train on your content is limited to foundation models, while the Privacy Statement still says conversation data is used to improve Copilot. “We also use your conversation data to monitor performance, troubleshoot and fix bugs and issues, prevent abuse, and to provide and improve Microsoft Copilot.” (Microsoft Privacy Statement)
- Microsoft's health data notice says health data may be used for advertising or marketing where you are given a choice or asked for consent. “We may use consumer health data for other purposes for which we give you choices and/or obtain your consent as required by law – for example, for advertising or marketing purposes.” (Microsoft Consumer Health Data Privacy Policy)
- For Copilot Health, a separate health space for paying Microsoft 365 subscribers that was temporarily unavailable when we checked, Microsoft says HIPAA does not apply to most products of its kind. “Copilot Health is a direct-to-consumer wellness product, and HIPAA does not apply to most direct-to-consumer wellness products. However, the information and data you provide to Copilot Health is protected by Microsoft's security policies and practices and is not used to train AI models.” (About Copilot Health (FAQ))
Limits of what we checked
- Older rules still apply to people using the older, un-updated Copilot app: Microsoft's older Privacy FAQ says conversations and uploaded files are used for AI training by default with an opt-out, are stored for 18 months by default, and may be read by human reviewers with no opt-out. That FAQ says content not moved to the updated app by 2026-12-31 will be permanently deleted. The cells above describe the updated app only.
- The new app's pages neither repeat nor withdraw the older 18-month storage period, so whether it still applies is not stated.
- Work and school accounts and Microsoft 365 Copilot for organizations run under different terms and are not covered here.
- The Copilot Terms of Use, Microsoft Privacy Statement, Consumer Health Data Privacy Policy and Transparency Note were keyword-searched, not read end to end.
- Microsoft's support pages show no date. The Privacy Statement ("Last Updated: November 2026") and the Consumer Health Data Privacy Policy ("Last Updated: July 2026") print a month only, so those cells have no page date.
- Whether Microsoft sells Copilot data was not established for the ordinary app: the U.S. State Data Privacy Notice was not read. Only Copilot Health has an explicit no-sale statement.
- No page says in so many words that Microsoft 365 subscribers' content is never used for ads; the pages say only that people without a subscription may see ads.
- The in-app settings screens could not be checked without an account.
Gemini (Google)
What we looked at: Consumer Gemini app on a personal Google Account, free and paid Google AI plans, United States
Used to train AI? On by default; can be turned off. Google's privacy page says saved chats and uploads are used to train its AI models. For adults the Keep Activity setting starts on. Turning it off, or using a Temporary Chat, stops this for future chats unless you send feedback.
“Google uses your activity to provide, develop, and improve its services (including training generative AI models), as well as to protect Google, its users, and the public with the help of human reviewers.”
Gemini Apps Privacy Hub, dated October 26, 2026; downloaded from the company’s site
How long is it kept? 18 months default; reviewed chats 3 years. Google's privacy page says chats are kept 18 months by default (you can choose 3 or 36 months, or no limit) and 72 hours with Keep Activity off. Chats read by human reviewers are kept up to three years even after you delete them.
“Chats reviewed by human reviewers (and related data like your language, device type, location info, or feedback) are not deleted when you delete your activity. Instead, they are retained for up to three years.”
Gemini Apps Privacy Hub, dated October 26, 2026; downloaded from the company’s site
Can people read it? Yes; reviewers include contractors. Google's privacy page says some chats are read by human reviewers, including outside companies working for Google. Turning Keep Activity off stops review for product improvement but not for safety. Reviewed chats are kept up to three years.
“A subset of chats are reviewed by human reviewers (including Google’s trained service providers) to help improve Google services.”
Gemini Apps Privacy Hub, dated October 26, 2026; downloaded from the company’s site
Used for ads or sold? Not used for ads; not sold. Google's privacy page says Gemini chats are not used to show you ads and that it would tell you if that changed. It also says Google does not sell your personal information. The ads sentence does not name uploads.
“Your Gemini Apps chats are not being used to show you ads. If this changes, we will clearly communicate it to you.”
Gemini Apps Privacy Hub, dated October 26, 2026; downloaded from the company’s site
What it says about HIPAA No statement found. We found no mention of HIPAA, the US health privacy law, in the Gemini Apps Privacy Hub, the Gemini activity and upload help pages, or the Google Privacy Policy. Google's only consumer HIPAA statement is about the separate Google Health app.
Health data handled differently? No health mode; general rules only. The Gemini app has no separate handling for health uploads. Google's general privacy policy has a clause on Washington and Nevada health data laws and says it does not personalize ads on health. Whether an uploaded lab report counts under that clause is not stated.
“The Washington My Health My Data Act and Nevada Senate Bill 370 provide specific rights in addition to those rights above, that apply to health information.”
Google Privacy Policy, dated October 16, 2026; downloaded from the company’s site
Also worth knowing
- Google itself tells users not to enter anything they would not want a reviewer to see. “Please don’t enter confidential information that you wouldn’t want a reviewer to see or Google to use to improve our services, including machine-learning technologies.” (Gemini Apps Privacy Hub)
- Deleting your activity does not delete chats that outside reviewers have already seen; those are kept for up to three years. “If you delete your activity, past chats already reviewed by service providers aren’t deleted because they aren’t connected to your Google Account. Instead, they’re retained for up to 3 years.” (Gemini Apps Privacy Hub)
- Even with Keep Activity off or in a Temporary Chat, chats can still be seen by human reviewers for safety purposes. “Even if your Keep Activity setting is off or you use temporary chats, Google still uses your chats to respond to you and help protect Google, our users, and the public, including with help from human reviewers.” (Gemini Apps Privacy Hub)
- The Gemini settings do not stop chats being turned into anonymized data used to improve Google services. “Your Gemini settings don’t control processing of your chats to create anonymized data to improve Google services.” (Gemini Apps Privacy Hub)
- The same lab report is treated differently in Google's separate Google Health app, where AI conversations are not used for training unless you opt in. “Your personal interactions with AI in the Google Health app are not used for training the underlying generative models, unless you specifically opt in under a separate agreement to use your data for research and development.” (FAQs on privacy: Google Health)
Limits of what we checked
- Google Workspace, education and developer (API) accounts run under different terms and are not covered here. Google's HIPAA material sits on Workspace pages, which were not read and do not describe the consumer app.
- The Gemini Apps Privacy Hub prints two dates, October 13, 2026 on its Privacy Notice section and October 26, 2026 on its questions section. The evidence file does not record which section each quote comes from; the page date shown is the later one.
- No page says in so many words that deleting a chat deletes the uploaded file; that is an inference from uploads being saved as part of your activity.
- How long deletion takes (about 2 months, with backups up to 6 months) comes from Google's general retention page, not from a Gemini page.
- The default-on setting is stated for users aged 18 or over, on a help page that shows no date.
- The Temporary Chat help page returned no readable text, and the Google Terms of Service and Gemini additional terms were not checked.
- Quotes were read from the US English pages; settings and defaults may differ in Europe and the UK.
- Google Health is a separate product from the Gemini app with its own rules; Google's page says medical records sent or uploaded to it are no longer protected by HIPAA.
Grok (xAI (now SpaceXSI))
What we looked at: grok.com and the Grok iOS and Android apps, individual users, United States; Grok inside X not covered
Used to train AI? May be used; opt-out when logged in. The company's FAQ says what you submit may be used to train its models. Logged-in users can turn off "Improve the model" or use Private Chat; the opt-out covers new conversations only, and feedback you send may still be used.
“We may use your content and interactions with Grok (e.g., prompts, searches, and other materials you submit) along with Grok's responses to train our models.”
Consumer FAQs, dated May 12, 2025; downloaded from the company’s site
How long is it kept? Until you delete; then 30 days. The policy says chats stay as long as you keep them, and that deleted conversations, Private Chats and deleted accounts are removed within 30 days unless needed for legal, compliance or safety reasons. Uploaded files are not addressed separately.
“Further, if you choose to delete any or all of your conversations or if you choose to delete your account, we will delete the data within 30 days unless it is necessary to retain the data for legal, compliance, or safety purposes.”
Privacy Policy, dated October 27, 2026; downloaded from the company’s site
Can people read it? Yes; authorized staff may review chats. The company's FAQ says a limited number of authorized personnel may review conversations, including to improve the model. No way to opt out is described, and the pages do not say whether these people include contractors.
“A limited number of our authorized personnel may review your conversations with Grok for specific business purposes, including improving model performance, investigating security incidents and potential misuse of our services, and complying with our legal obligations.”
Consumer FAQs, dated May 12, 2025; downloaded from the company’s site
Used for ads or sold? Not sold; ad use not addressed. The company's FAQ says it does not sell your data. The pages neither say nor deny that chats or uploads are used for advertising; the advertising wording covers only cookies on its websites.
“No. We do not sell your data to third parties as defined by applicable laws.”
Consumer FAQs, dated May 12, 2025; downloaded from the company’s site
What it says about HIPAA No statement found. We found no mention of HIPAA, the US health privacy law, in the consumer Privacy Policy, Consumer FAQ, Terms of Service or Acceptable Use Policy. The agreement hospitals and clinics need is offered only to business customers of the developer service.
Health data handled differently? Asks users not to share health data. The privacy policy says the company does not aim to collect sensitive information such as health details and asks you not to provide it. No health mode and no consumer health data notice were found.
“We do not aim to collect sensitive personal information (ex., information related to racial or ethnic origin, political opinions, religion or other beliefs, health, biometric scans, criminal background, or trade union membership) and ask that you do not provide us with any such information.”
Privacy Policy, dated October 27, 2026; downloaded from the company’s site
Also worth knowing
- The same FAQ that asks users not to share sensitive information invites them to upload documents for analysis. “You can also upload your documents and ask Grok to analyze them” (Consumer FAQs)
- The 30-day deletion promise does not cover conversations that have been stripped of your identity and separated from your account. “If you delete conversations from your account or if you use Private Chat, conversations will be removed from our systems within 30 days, unless they have been de-identified or pseudoanonymized and disassociated from your account or we have to retain them for safety, security, or legal reasons.” (Consumer FAQs)
- People who use Grok without logging in cannot opt out of training in some regions. “As a result, in some regions (excluding the EU/UK), when you use Grok without logging in, you won’t have the option to opt out of model training.” (Consumer FAQs)
- The company offers a HIPAA agreement only for its developer service, not for the consumer apps. “This form is for customers requesting a Business Associate Agreement (BAA) from SpaceXSI to enable HIPAA compliance when using SpaceXSI APIs.” (BAA questionnaire)
- grok.com uses advertising cookies that let ad partners show you ads on other sites. “Our third-party advertising partners may use these cookies to show you relevant ads on other third-party websites and platforms.” (Cookie Policy)
Limits of what we checked
- Grok used inside X (formerly Twitter) falls under X's own privacy policy and was not audited.
- The company's legal pages now use the name SpaceX Super Intelligence LLC ("SpaceXSI"), though they are still served from x.ai.
- Business and developer (API) customers have separate terms, including the HIPAA agreement, which are not covered here.
- The FAQ does not use the word "default" or say whether the "Improve the model" switch starts on or off; that training is on unless you opt out is inferred from its wording.
- The Consumer FAQ is dated May 12, 2025, older than the Privacy Policy and Terms of Service (October 27, 2026).
- No page says whether deleting a chat deletes the uploaded file, or whether turning off training or using Private Chat stops staff review.
- The policy allows disclosure to "related companies" without defining the term.
- No separate Grok help centre could be reached, the Acceptable Use Policy was keyword-searched only, and the Europe privacy addendum was not read.
Meta AI (Meta)
What we looked at: Meta AI assistant in the Meta AI app, WhatsApp, Instagram and Messenger, United States
Used to train AI? Used to improve AI; no US opt-out found. Meta's pages say your interactions with Meta AI, including messages and replies, are used to train and improve its AI. We found no statement of an opt-out or setting for US users; Meta's objection help page describes that right for the EU and UK only.
“For example, when you chat with AI, we may use those messages and replies to improve the model.”
How generative AI models work (Meta Privacy Center); read from an Internet Archive copy
How long is it kept? No fixed period; deletion may leave Meta's copy. The Privacy Policy says information is kept as long as Meta needs it, decided case by case, and deletion can take up to 90 days plus 90 more for backups. The AI Terms say deleting messages, threads or your account may not delete Meta's copy.
“Depending on how you use AIs, deleting individual messages, entire threads, or even your account may not delete our copy of your personal information, and the Privacy Center describes the correct way to delete your personal information.”
Meta AI Terms, dated May 13, 2026; read from an Internet Archive copy
Can people read it? Yes: human review, including outside vendors. The AI Terms say Meta will sometimes review conversations with its AIs, by automated or human review, and may use third-party vendors. Uploaded documents count as content under those terms. Chats using "Private Processing" are the stated exception.
“In some cases, Meta will review your interactions with AIs, including the content of your conversations with or messages to AIs, and this review may be automated or manual (human).”
Meta AI Terms, dated May 13, 2026; read from an Internet Archive copy
Used for ads or sold? AI conversations personalize ads; not sold. A Meta announcement says interactions with Meta AI are used to personalize ads and content from December 16, 2025 in most regions, and that sensitive topics such as health are not used to show ads. The Privacy Policy says Meta does not sell your information.
“We’ll start to use your AI interactions to personalize content and ads recommendations after these changes go into effect on December 16. We’re rolling out these changes in most regions, and we hope to offer these more personalized experiences everywhere soon.”
Meta Newsroom post on AI interactions and ads (October 2025), dated October 1, 2025; read from an Internet Archive copy
What it says about HIPAA No statement found. We searched the full text of 17 archived Meta pages for "HIPAA", including the Privacy Policy, the AI Terms, the Consumer Health Data policy and the Meta AI help pages, and found no mention.
Health data handled differently? No health mode; warns against sharing sensitive topics. No separate health feature was found. The AI Terms tell users not to share sensitive information they do not want the AI to keep and use. The Privacy Policy says health information is not used for ads but is used to improve products.
“Do not share information that you don’t want the AIs to use and retain, such as information about sensitive topics.”
Meta AI Terms, dated May 13, 2026; read from an Internet Archive copy
Also worth knowing
- We found no statement that US users can opt out of having Meta AI conversations used to improve AI; Meta's objection help page describes the right to object for the EU and UK only. “Under the General Data Protection Regulation (GDPR), if you’re in the EU or the UK, you have the right to object at any time to certain types of processing of your personal information.” (Meta help page on objections)
- Meta says health and other sensitive topics in Meta AI conversations are not used to show ads; this exclusion covers ads only, not AI improvement or human review. “When people have conversations with Meta AI about topics such as their religious views, sexual orientation, political views, health, racial or ethnic origin, philosophical beliefs, or trade union membership, as always, we don’t use those topics to show them ads.” (Meta Newsroom post on AI interactions and ads (October 2025))
- Meta promotes Meta AI for health questions, including images and charts, while its AI Terms tell users not to share information about sensitive topics. “With Muse Spark, Meta AI is now able to help you navigate health questions with more detailed responses, including some questions involving images and charts.” (Meta Newsroom post introducing Muse Spark (April 2026))
- When Meta AI cannot answer, Meta says it sends the messages you wrote to outside partners such as search engines. “If the AIs cannot answer your query, Meta shares the messages you send to AI and general information, like your region, with select partners so you can get better results.” (Remove posts, chats and media from Meta AI and Vibes (Meta Help Center))
- Meta says deleting information does not change a model that has already learned from it; the deleted information is only kept out of future training. “If you delete a piece of information a model learned from (like a public post), the model won’t change in that moment. However, your information itself will be deleted, and once it’s deleted, it won’t be used for future training.” (How generative AI models work (Meta Privacy Center))
Limits of what we checked
- Meta's live pages could not be downloaded, so quotes come from Wayback Machine copies, and the source links point to those archived copies. Most are 1 to 30 days older than our check on 2026-11-09; the objection help page copy is from 2026-05-21.
- Archived copies show what Meta served to the Internet Archive's crawler, whose location is not recorded, and Meta's help pages contain blocks shown only in some regions. A sentence offering an objection to AI training exists in one help page's data but was not displayed in the archived copy or in a copy saved from a browser on 2026-11-10 (browser location not recorded, one of four tabs captured).
- The missing US opt-out is an absence in the pages we read, not a statement by Meta that US users cannot object. The objection form itself was not read.
- The 90-day deletion periods come from Meta's general Privacy Policy. No page says that deleting Meta AI chats counts as a deletion request under that policy, and the deletion help page does not mention uploaded files or Meta's own copy.
- Meta says AI interactions personalize ads "in most regions" without listing them. No page says how a health document uploaded to a chat is classified for the ad exclusion.
- WhatsApp's help pages on Meta AI could not be verified against an archived copy and are not quoted. No Meta AI help page about uploading documents was found, and Instagram- and Messenger-specific help pages were not checked.
- The Privacy Center pages and help pages show no date or only a relative date, so their page dates are left blank.
- Meta's business AI tools, developer API and its separate Muse agent are sold or offered under other terms and are not covered here.
Perplexity (Perplexity)
What we looked at: Free, Pro and Max consumer plans, United States
Used to train AI? On by default; can be turned off. A help page says the "AI data retention" setting is on by default for Free, Pro and Max users and can be switched off. Another help page says uploaded files are used only to answer you, but the Privacy Notice has no such exception.
“AI model training: Some data may be used to train AI models and improve search quality, unless you opt out through Account Settings”
Data Collection at Perplexity, dated August 31, 2026; saved from a browser as a PDF
How long is it kept? Files 30 days; chats while account active. Help pages say files attached to a chat are kept 30 days, though answers already drawn from them stay in the chat, and deleting a chat deletes its files. Other personal information is kept while your account is active.
“Perplexity doesn't store files indefinitely—uploaded files are retained for 30 days (7 days for Enterprise Pro users), allowing for follow-ups if needed. After this period, files are automatically removed from our system, but follow-up responses will still have context from previous questions in the conversation.”
Typing a question vs. uploading a file or image; saved from a browser as a PDF
Can people read it? No statement found. No saved page says whether Perplexity staff or contractors can read chats or uploads. We checked the Terms of Service, Privacy Notice, seven help pages and the two Perplexity Health pages. The Terms reserve a right to access and "screen" content without saying who does.
Used for ads or sold? Not sold for money; some data to advertisers. The Privacy Notice says data is not sold "for money" but lists advertising partners as recipients of identifiers, search history and approximate location for targeted ads. Health-related information is marked as not shared for advertising. A help page says data goes only to service providers.
“We may share data with third parties to measure the performance of our advertising, including via server-to-server connections and web attribution and measurement technologies (e.g., a cookie or a pixel). The third party may also collect data over time and across websites operated by other companies.”
Perplexity Privacy Notice, dated October 23, 2026; saved from a browser as a PDF
What it says about HIPAA No statement found. Neither the Privacy Notice nor the Terms of Service mentions HIPAA, and no saved page addresses it for ordinary use. The only statement is on the Perplexity Health help page, which says HIPAA does not apply to consumer health products like Perplexity Health.
Health data handled differently? Health data notice; separate Perplexity Health feature. Perplexity Health, a feature for US Pro and Max subscribers, says health chats and files are not used to train AI models. The Privacy Notice has a health data paragraph covering anything health-related you type or upload, but makes no such promise for ordinary chats.
“Is my health data used to train AI models? No. Your health chats, connector data, files, and memories are not used to train AI models and will never be sold to third parties.”
What is Perplexity Health?, dated October 14, 2026; saved from a browser as a PDF
Also worth knowing
- A help page says uploaded files are used only to answer you, but the Privacy Notice and Terms of Service, the legal documents, allow uploads to be used to improve services and AI models with no exception for files. “Your files stay private. We use their content only to customize responses to your questions and tasks. They're never accessed or shared for other purposes unless you share a session.” (Security and Privacy with File Uploads)
- A help page says information goes only to service providers, while the newer Privacy Notice lists advertising partners as recipients of identifiers, search history and location. “Data privacy commitment: Perplexity does not sell your data. Information is only shared with trusted service providers (payment processors, customer support) or when required by law.” (Data Collection at Perplexity)
- Perplexity's own health feature states that HIPAA does not apply to it. “Perplexity Health is designed for personal wellness. HIPAA does not apply to consumer health products like Perplexity Health.” (What is Perplexity Health?)
- Help pages say personal data is removed within 30 days of deleting an account, but the Terms of Service say the company is not obliged to delete your content. “In the event of Account deletion for any reason, the Company may, but is not obligated to, delete any of Your Content.” (Perplexity Terms of Service)
- Right after saying health data is not used to train AI models, the Perplexity Health page adds that questions and answers may be used to improve Perplexity, without explaining the difference. “We also may use your questions and answers to help improve Perplexity, as explained in our Privacy Notice.” (What is Perplexity Health?)
Limits of what we checked
- Perplexity's site blocked automated access, so every quote comes from PDFs of the live pages saved from a browser on 2026-11-09 or 2026-11-10.
- The PDFs' text was read by optical character recognition and matched allowing for differences in quotation-mark and dash style, so a quote may show a hyphen or straight quote where the page has a dash or curly quote. Quotes should be checked by eye against the PDFs before they are relied on.
- The saved PDFs do not print their web addresses; source links are the addresses a search returned for the same page titles and were not confirmed. The help page "Typing a question vs. uploading a file or image" shows no date.
- The table in the Privacy Notice listing advertising partners was read by eye, and reading its third column as "sharing for targeted advertising" is an inference from the page layout. The notice does not define "Advertising Partners" or say whether health text typed as a query counts as health information or as search history.
- Three passages of the Perplexity Health help page were hidden at page breaks in the saved PDF and are not quoted.
- No statement was found on how long an individually deleted chat or file remains in Perplexity's systems or backups; the only timed commitment is account deletion within 30 days. The pages call chats "sessions" or "threads".
- Perplexity Health is available only to Pro and Max subscribers in the US. The pages do not say whether a file attached to an ordinary chat that triggers Perplexity Health is treated as a health file.
- Enterprise and API products are sold under other terms and are not covered here.
Training is usually on until you turn it off
OpenAI, Anthropic, Google, xAI, and Perplexity each describe a setting that stops new chats being used for training. The setting matters less than it sounds in two ways. First, it tends to cover future chats only. Second, sending feedback can override it: OpenAI’s help page says: “For individual plans (Free, Plus, and Pro), if you submit feedback on a response—for example, by selecting thumbs up or thumbs down—we may use the entire conversation associated with that feedback to improve our models, even if you have opted out of model training.”
Microsoft says Copilot prompts and files “aren’t used to train foundation models,” a narrower promise than it first appears, since its privacy statement also says conversation data is used to “provide and improve Microsoft Copilot.” For Meta AI we found no opt-out for US users; the help page describing a right to object addresses people in the EU and UK.
Deleting a chat is not the same as deleting the report
This is the finding we would most want a patient to know before uploading.
- ChatGPT saves uploaded files to a Library. Its help page says: “Deleting a chat containing files does not delete a separate file that remains saved in Library. To remove a saved Library file, delete it from Library.” It also keeps memories apart from chats: “Saved memories are stored separately from chat history. Deleting the original chat does not automatically delete a separate saved memory.” Removing a report can therefore take three deletions: the chat, the Library file, and any saved memory.
- Gemini: “Chats reviewed by human reviewers (and related data like your language, device type, location info, or feedback) are not deleted when you delete your activity. Instead, they are retained for up to three years.”
- Meta AI: “Depending on how you use AIs, deleting individual messages, entire threads, or even your account may not delete our copy of your personal information, and the Privacy Center describes the correct way to delete your personal information.”
- Claude and Grok both state a 30-day limit for removing deleted conversations from their systems, with exceptions each describes.
The companies tell you not to send sensitive information
Five of the seven say so in their own pages, which sits oddly beside products that invite you to upload documents:
- OpenAI: “Please do not enter sensitive information that you would not want reviewed or used.”
- Google: “Please don’t enter confidential information that you wouldn’t want a reviewer to see or Google to use to improve our services, including machine-learning technologies.”
- Microsoft: “Copilot may include both automated and manual (human) processing of data. You shouldn’t share any information with Copilot that you don’t want us to review.”
- Meta: “Do not share information that you don’t want the AIs to use and retain, such as information about sensitive topics.”
- xAI: “We do not aim to collect sensitive personal information (ex., information related to racial or ethnic origin, political opinions, religion or other beliefs, health, biometric scans, criminal background, or trade union membership) and ask that you do not provide us with any such information.”
Advertising
Meta says conversations with Meta AI are used to personalize ads, with health and other sensitive topics excluded from that use. Microsoft says people without a Microsoft 365 subscription may see ads that “relate to the current interaction,” and names “files you share” in the same sentence. OpenAI shows ads on its Free and Go plans based on the current chat and says they are not eligible to appear near personal health topics. Perplexity’s privacy notice lists advertising partners among the recipients of some data, while one of its own help pages says information goes only to service providers. Google and Anthropic say chats are not used for ads.
The separate health features
OpenAI, Perplexity, and Microsoft each offer a health space with stricter rules than the ordinary chat, typically a promise not to train on what you put there. Two points are easy to miss. A lab report dropped into an ordinary chat does not get those rules. And the health spaces are not under HIPAA either: OpenAI says its Health feature “is not intended for clinical or covered-entity use and does not offer a Business Associate Agreement.”
The lab-result apps
These are consumer products that accept an uploaded or photographed lab report and return an AI interpretation. There are far more of them than anyone could audit, so we chose by rule and not by reputation. 7 of the 12 advertise HIPAA compliance, and the table starts there.
| App | HIPAA claim in marketing | What its policy says about HIPAA | Used to train AI? | Whose AI reads the report? | How long is it kept? |
|---|---|---|---|---|---|
| AI Lab Results Analyzer | None found | No statement found | No statement found | No statement found | No statement found |
| Blood Buddy | None found | No statement found | Anonymised results used to improve AI | No statement found | Kept until you ask for deletion |
| BloodGPT | Yes, as a badge | Named as applicable requirement; no role stated | Home page says never used for training | Three models; diagram shows Claude, Gemini, GPT-4o | No period given; deletion by email request |
| BloodTracker AI | None found | One audit mention; no role stated | Says OpenAI's terms prohibit training | OpenAI | No fixed period; deletion by email request |
| Docus | Yes | Says HIPAA may not apply; complies voluntarily | No statement found | No statement found | No fixed period; deletion on request |
| Kantesti AI Blood Test Analyzer | Yes | Aims to comply where applicable; role unstated | Terms: anonymized data may improve AI | Described as proprietary; no outside provider named | Free: up to 90 days; paid: while subscribed |
| Levels | Yes | Says it is not a Covered Entity | Providers don't train; improvement opt-out by email | OpenAI, Anthropic, Google and xAI | Kept indefinitely unless deletion requested |
| NoctorAI / Visualizer | Yes | No statement found | No statement found | No statement found | Up to 12 months after account goes idle |
| SelfDecode | Yes | No statement found | Says no AI training on your information | Unnamed outside model, run in SelfDecode's cloud | While account open; in-app deletion within one day |
| SiPhox Health (upload analyzer, "Sai") | None found | No statement found | Policy allows model training; no opt-out found | OpenAI and ChatGPT named as examples | No time limit stated; deletion offered |
| SnapLabs | None found | No statement found | Says not used to train AI models | Google (Gemini 3.1 Flash-Lite) | On device by default; deletable in app |
| Wizey | Yes | No statement found | No statement found | No statement found | As long as necessary; deletion offered |
AI Lab Results Analyzer (COCONUT AI LLP)
What we looked at: iOS app, US App Store listing (version 1.5); the app's privacy policy could not be read
Used to train AI? No statement found. The app's privacy policy could not be read, so we looked only at its App Store listing. It says nothing about using uploads to train or improve AI, and mentions no opt-out.
How long is it kept? No statement found. The app's privacy policy could not be read, so we looked only at its App Store listing. It says information is encrypted and stored securely, but gives no storage period, no storage location and no way to delete.
Can people read it? No statement found. The app's privacy policy could not be read, so we looked only at its App Store listing. It says nothing about whether staff or contractors can see uploaded reports.
Used for ads or sold? No statement found. The app's privacy policy could not be read, so we looked only at its App Store listing. Its text says nothing about sharing, selling or advertising. The developer's privacy label shows no tracking section, which is not the same as a promise not to share.
What it says about HIPAA No statement found. The app's privacy policy could not be read, so we looked only at its App Store listing. The word HIPAA does not appear there, and nothing is said about being a covered entity or business associate.
Health data handled differently? No statement found. The app's privacy policy could not be read, so we looked only at its App Store listing. It calls health data sensitive in one general sentence but describes no special rules, health privacy notice or removal of identifiers.
Whose AI reads the report? No statement found. The app's privacy policy could not be read, so we looked only at its App Store listing. No AI model or company is named.
Privacy label on the App Store: Data Not Linked to You: Health, Photos or Videos (purpose: Product Personalization).
Also worth knowing
- The privacy policy link on the App Store listing returned no readable policy, and the copy saved from a browser turned out to be the privacy policy of Notion, the hosting service, not the app's.
- The developer's App Store label says the health data and photos it collects are not linked to your identity, although a lab report normally carries the patient's name; the listing does not explain. “The following data, which may be collected but is not linked to your identity, may be used for the following purposes:” (App Store listing, App Privacy)
- The listing says information is stored, so uploads are not simply processed and discarded, but it does not say where or for how long. “All information is encrypted and stored securely, ensuring a safe and private experience.” (App Store listing, description)
- The listing names a company as the seller and an individual as the copyright holder, and gives no developer website, address or country.
Limits of what we checked
- The app's privacy policy could not be read. The App Store links to a Notion page that returned no content when downloaded, and the PDF saved from a browser for that link was Notion's own corporate privacy policy, which says nothing about this app.
- Every cell is therefore "No statement found". That describes what we could read, not what the company's policy may say.
- The only official page read was the US App Store listing (latest version entry 1.5, dated 12/20/2025). The privacy label is reported by the developer and not verified by Apple.
- There are no developer-specific terms; the listing uses Apple's standard licence agreement. The support link returned no content either.
- A Google Play app with the same name comes from a different publisher and was not examined.
Blood Buddy (Dillon Prime Limited)
What we looked at: The Blood Buddy app and mybloodbuddy.com
Used to train AI? Anonymised results used to improve AI. The policy says anonymised text from blood test results is used to improve the AI and models in the app. No opt-out was found other than deleting your account. The policy does not say whose models these are.
“We use aggregated and anonymized usage data, including text data from the anonymized blood test results, to understand how users interact with the platform, to improve the AI and models in the app, and to identify areas for improvement.”
Privacy Policy, dated January 7, 2025; downloaded from the company’s site
How long is it kept? Kept until you ask for deletion. The policy sets no time limit: blood test data stays unless UK law requires removal or you ask. It says deleting your account erases everything immediately and permanently. The company's pages disagree on whether backups are kept.
“Blood test data is kept to ensure the product can improve and continue to function. This will stay in place unless we are required to remove it by UK law or the user requests it.”
Privacy Policy, dated January 7, 2025; downloaded from the company’s site
Can people read it? Says developers have no access. The policy says Blood Buddy developers have zero access to user data. The same policy says access to its servers is restricted to authorized personnel, without saying who they are or what they can see.
“Blood Buddy developers have zero access to user data at any time.”
Privacy Policy, dated January 7, 2025; downloaded from the company’s site
Used for ads or sold? No advertising use without your consent. The policy says your information is not used for marketing or advertising without your explicit consent. A separate security page says data is not sold, shared or monetized; the privacy policy itself has no sentence saying data is not sold.
“We do not use your information for marketing, advertising, or any other commercial purposes without your explicit consent.”
Privacy Policy, dated January 7, 2025; downloaded from the company’s site
What it says about HIPAA No statement found. HIPAA is not mentioned on the privacy policy, terms, security page, home page, contact page or App Store listing. The company makes no HIPAA claim and frames your rights under United Kingdom data protection law.
Health data handled differently? Says only de-identified text reaches its servers. The policy says most processing happens on your device and only text with identifying details removed is sent to its servers. It admits no such technique is perfect. No Washington-style consumer health data notice or separate sensitive-data section was found.
“Only de-identified data in the form of text is transmitted to our servers.”
Privacy Policy, dated January 7, 2025; downloaded from the company’s site
Whose AI reads the report? No statement found. No AI model or outside AI company is named on the privacy policy, terms, security page or home page. The pages do not say whether the AI runs on your device, on Blood Buddy's servers or at another company.
Privacy label on the App Store: Data Not Linked to You: Identifiers (User ID) and Diagnostics; no Health category declared.
Also worth knowing
- The security page says the company keeps no backups, which the terms of use contradict. “If you disconnect all synced devices, your data is gone forever (we do not keep backups).” (Security page)
- The terms of use say daily backups are made, and no page says how long backups last after you delete your data. “we perform daily backups of our database and server files as part of our routine maintenance and disaster recovery procedures.” (Terms of Use)
- The terms say data derived from your results, once de-identified or combined, belongs to the company. “While you retain ownership of your original content, the derivatives created from your content, such as de-identified or aggregated data, belong to us.” (Terms of Use)
- The policy admits that removing identifying details cannot fully rule out someone being identified again. “While we take extensive measures to de-identify your data, please be aware that no de-identification technique is perfect, and there is always a theoretical risk of re-identification.” (Privacy Policy)
- The App Store privacy label declares no Health category, while the privacy policy says anonymised blood test results are stored on the company's servers. “We collect and store only the anonymized blood test results, markers, and their values in the form of text.” (Privacy Policy)
Limits of what we checked
- The terms of use name Dillon Prime Limited (England and Wales) as the company, the App Store listing names Hand Clicked Limited as the seller, and the privacy policy names no legal entity.
- The statement that data is not sold, shared or monetized is on an undated security page, not in the privacy policy.
- The policy says data processed by AI models is not retained after processing, and also that anonymised results are stored on its servers; it does not explain the difference.
- The security page says no account or email is required, while the policy's deletion steps begin with logging in to an account; the two pages may describe different versions of the product.
- The privacy policy and terms were last updated on 7 January 2025.
- The signed-in web app and the Google Play listing were not checked.
BloodGPT (BloodGPT Limited)
What we looked at: Home page and privacy policy as saved on 2026-11-10; the page for individual plans was not saved
HIPAA claim in its marketing (a badge image, not a sentence; Home page: an image of a badge, shown three times (logo row under the product screenshot, the security block, the footer). There is no HIPAA sentence in the home page text.): “HIPAA COMPLIANT”
Used to train AI? Home page says never used for training. The home page says your data is never used to train models. The privacy policy does not repeat that promise and says nothing about training either way. Whether the outside AI companies may train on the data is not stated.
“No sharing. Your data is never used to train models.”
Home page; saved from a browser as a PDF
How long is it kept? No period given; deletion by email request. The privacy policy says uploaded files are kept only as long as needed to analyse them, then deleted or anonymized on a schedule it does not publish. You can ask for erasure by email. No time limit for deletion is given.
“Blood Test Files: Uploaded files are stored only for the duration necessary to analyze and provide the explanations. We have a defined retention schedule after which test results are automatically deleted or anonymized, unless we are obligated by law to store them longer.”
Privacy Policy, "Data Retention", dated April 30, 2025; saved from a browser as a PDF
Can people read it? No statement found. We looked in the saved home page and privacy policy. The policy mentions restricted access and staff privacy training, but neither page says whether staff or contractors can view uploaded reports.
Used for ads or sold? No advertising without consent; sale not addressed. The privacy policy says personal data is not used for advertising or marketing unless you explicitly agree. It also says data may be shared with service providers such as hosting and analytics companies. Neither saved page mentions selling data.
“We do not use your Personal Data for advertising or marketing unless you have explicitly consented to such use.”
Privacy Policy, dated April 30, 2025; saved from a browser as a PDF
What it says about HIPAA Named as applicable requirement; no role stated. The privacy policy mentions HIPAA once, as one of the privacy requirements its data protection officer enforces. It does not say the company is a covered entity or business associate under that law, and no HIPAA contract is offered.
“The DPO is responsible for monitoring and enforcing compliance with applicable privacy requirements, including but not limited to the GDPR, HIPAA, and CPRA.”
Privacy Policy, "Data Protection Officer", dated April 30, 2025; saved from a browser as a PDF
Health data handled differently? Processed on explicit consent under EU law. The privacy policy says uploads can contain sensitive health data and that by providing it you explicitly consent to its processing under European data protection law (GDPR). No US state health-data notice was found.
“Explicit Consent (Article 9(2) GDPR): When you provide us with your health data for analysis, you explicitly consent to our processing of that data for the purpose of delivering the Service.”
Privacy Policy, dated April 30, 2025; saved from a browser as a PDF
Whose AI reads the report? Three models; diagram shows Claude, Gemini, GPT-4o. The home page says the system runs on three AI language models. A diagram beside that sentence shows the logos of Claude, Gemini and GPT-4o. No sentence says which company receives uploaded reports, and the privacy policy names none.
“A custom multi-agent system powered by three specialised LLMs, with a deterministic pipeline for value extraction and threshold logic — same input, same output.”
Home page, "Not another GPT solution"; saved from a browser as a PDF
Also worth knowing
- The home page shows a "HIPAA COMPLIANT" badge three times, but the privacy policy never says the service is HIPAA compliant; its only mention of HIPAA is this sentence about the data protection officer. “The DPO is responsible for monitoring and enforcing compliance with applicable privacy requirements, including but not limited to the GDPR, HIPAA, and CPRA.” (Privacy Policy, "Data Protection Officer")
- The home page says "No sharing", while the privacy policy says data may be shared with outside service providers and does not exclude health data. “Third-party Service Providers: We may share certain data with third-party service providers who help us operate the Service (e.g., cloud hosting, analytics).” (Privacy Policy, "Data Sharing and Transfers")
- The published privacy policy contains what reads as a drafting note left in square brackets, in the part about consent for advertising and marketing. “Consent (Article 6(1)(a) GDPR): [only if applicable: data from non-essential cookies and other tracking technologies, and data to conduct advertising and marketing]” (Privacy Policy, legal bases)
- The policy says uploads are kept only for as long as needed to analyse them, while the home page advertises tracking results over time; the policy does not explain how both hold. “Track Biomarker Trends Over Time” (Home page)
- The privacy policy's cookie section says the site uses advertising cookies that follow your browsing. “These cookies track your browsing activity to deliver relevant advertisements and measure advertising campaign effectiveness.” (Privacy Policy, "Marketing/Advertising Cookies")
Limits of what we checked
- Only two pages were read, both from PDFs saved from a browser on 2026-11-10: the home page and the privacy policy (last updated 2025-04-30).
- The privacy policy PDF does not print its web address, so its source URL is left empty.
- The home page FAQ was saved collapsed. The answer to "Is BloodGPT compliant with data privacy regulations?" was not read and may contain a HIPAA sentence.
- The Terms of Use, Subprocessors list, Trust Center and the page for individual plans were not saved or read; they may say more about HIPAA, sharing and which AI companies receive data.
- The HIPAA badge and the diagram with the Claude, Gemini and GPT-4o logos are pictures and were read by eye. The diagram is an illustration, not a statement about where data goes.
- BloodGPT also sells to doctors, labs and clinics; the pages do not say whether the HIPAA badge refers to that business. No sentence about US availability was found; the company is in Cyprus.
- No app store privacy label is recorded for this product.
BloodTracker AI (WOLKK PTE. LTD. (App Store seller; the website names no company))
What we looked at: iOS app and website; privacy policy, terms and home page as saved on 2026-11-10
Used to train AI? Says OpenAI's terms prohibit training. The privacy policy says OpenAI, which processes the reports, is barred by its own terms from training on your data. It does not say whether BloodTracker AI uses uploads to improve its own systems. No opt-out was found.
“OpenAI processes this data under its own privacy policy and API data usage terms, which prohibit using your data to train their models.”
Privacy Policy, section 3, dated August 14, 2026; saved from a browser as a PDF
How long is it kept? No fixed period; deletion by email request. The privacy policy says data is kept only as long as necessary, then deleted or anonymized, with no period given. Deletion is a right you exercise by emailing support. The home page says data is stored on your device.
“We retain your data only as long as necessary for the purposes outlined in this policy or as required by law. After this period, data is securely deleted or anonymized.”
Privacy Policy, dated August 14, 2026; saved from a browser as a PDF
Can people read it? No statement found. We looked in the privacy policy, terms and home page. The policy says only that access is limited to authorized personnel; it does not say whether staff or contractors can view uploaded reports.
Used for ads or sold? Not sold; shared with OpenAI, Firebase, AppsFlyer. The privacy policy says personal and health data is not sold. It names three companies that receive data: OpenAI to process reports, Google Firebase for analytics, and AppsFlyer to measure app installs. No statement about advertising use was found.
“We do not sell your personal or health data. We may share information only in these cases:”
Privacy Policy, section 3, dated August 14, 2026; saved from a browser as a PDF
What it says about HIPAA One audit mention; no role stated. The privacy policy mentions HIPAA once, as something its security audits aim at. It does not say the company is covered by the law as a covered entity or business associate, and no HIPAA contract is mentioned.
“Regular Audits: To ensure compliance with GDPR, HIPAA, and other applicable regulations.”
Privacy Policy, section 4 "Data Security", dated August 14, 2026; saved from a browser as a PDF
Health data handled differently? No statement found. We looked in the privacy policy, terms and home page. The policy lists health data as something it collects but has no section on sensitive data and no US state health-data notice. It does not say identifiers are removed before reports go to OpenAI.
Whose AI reads the report? OpenAI. The privacy policy says report images and PDFs, the text read from them and the test values are sent to OpenAI for reading and analysis. The home page instead describes the reading step as on-device AI.
“With OpenAI, L.L.C., to process your blood test report images/PDFs, extracted text, and biomarker values for OCR and AI-generated analysis.”
Privacy Policy, section 3, dated August 14, 2026; saved from a browser as a PDF
Privacy label on the App Store: Data Not Collected. The company's privacy policy lists collection of name, date of birth, health data, device information and IP address, and sharing with OpenAI, Google Firebase and AppsFlyer.
Also worth knowing
- The home page says the app reads your results with on-device AI, while the privacy policy says report images and PDFs are sent to OpenAI for reading and analysis. “On-device AI extracts and standardizes all your test values.” (Home page, "How It Works")
- The home page says your data is stored on your own device; the policy says reports go to OpenAI, and no page says whether the company's servers keep a copy. “Your data is stored securely on your local device.” (Home page, privacy section)
- The home page promises no sharing without your explicit consent, but the privacy policy lists sharing with OpenAI, Google Firebase and AppsFlyer without making each depend on consent. “Used solely for providing and improving the app's functionality; we do not sell or share your data with third parties without your explicit consent.” (Home page, "Never Sold" card)
- The App Store privacy label says "Data Not Collected", while the company's own privacy policy lists health data, name, date of birth and device data as collected. “Health Data: Information you input or share, such as symptoms, medical history, or wellness data.” (Privacy Policy, list of data collected)
- The website names no company or country; the terms say only that the law of wherever the app "operates" applies. “These Terms shall be governed by and construed in accordance with the laws of the jurisdiction in which BloodTracker AI operates, without regard to conflict of law principles.” (Terms of Service)
Limits of what we checked
- The privacy policy, terms and home page were read from PDFs printed from a browser on 2026-11-10 and checked by eye and by machine text recognition; the PDFs' own text layer is fragmented.
- The App Store label ("Data Not Collected"), the seller name and the App Store description were read from the US listing on 2026-11-09 and were not re-checked on 2026-11-10. The label is reported by the developer.
- A few lines are hidden under the site's header at page breaks in the saved PDFs. Quotes confirmed only from the fragmented text layer, including the home page's "anonymized where possible" line, were not used.
- The website names no company or address. The company name shown here is the App Store seller.
- The same privacy policy covers a second app, BP Tracker. The Google Play listing and any in-app privacy text were not checked.
Docus (Docus, Inc.)
What we looked at: Consumer lab test interpretation and AI Doctor pages, privacy policy and terms; trust center Overview tab only
HIPAA claim in its marketing (a sentence; Lab test interpretation page, "Your data is secure" block): “Any data shared while using our services is safeguarded by advanced security protocols, ensuring compliance with HIPAA and GDPR.”
Used to train AI? No statement found. We looked in the privacy policy, terms and product pages. None mentions training AI on uploads or a way to opt out. The policy does allow use of your information to improve the services, and the terms allow use of anonymized medical information.
How long is it kept? No fixed period; deletion on request. The privacy policy says information is kept as long as necessary, with no period given. You can ask for deletion by email or delete your account. Nothing is said about deleting a single uploaded report.
“We will only keep your Personal information for as long as it is necessary for the purposes set out in this Privacy Policy unless a longer retention period is required or permitted by applicable law (such as tax, accounting, or other legal requirements).”
Privacy Policy, dated March 26, 2024; downloaded from the company’s site
Can people read it? Staff and contractors may access and review. The terms say you authorize the company's employees, contractors, medical experts and agents to access and review your medical information as needed to provide the service. The privacy policy adds that Docus may review records to check service quality.
“In addition, you authorize the Company and their respective Medical Experts, employees, contractors, business associates and/or agents to access, review, research, analyze, discuss and copy your Medical and Personal Information as may be necessary to provide the Services and the Reports you request.”
Terms of Use, dated March 26, 2024; downloaded from the company’s site
Used for ads or sold? Not sold; shared with marketing vendors. The privacy policy says information is not sold, traded or rented. It also says information may be shared with outside vendors, including advertising and retargeting companies, and with affiliates. It does not exclude medical information from that sharing.
“We do not sell, trade, or rent your Information to others.”
Privacy Policy, dated March 26, 2024; downloaded from the company’s site
What it says about HIPAA Says HIPAA may not apply; complies voluntarily. The privacy policy says HIPAA, the US health privacy law, may not apply to its services but that it has decided to follow it anyway. It does not say Docus is a covered entity or business associate, and no HIPAA contract is offered.
“Although our Services may not include the specific transactions to which HIPAA may apply, we have nonetheless decided to operate in compliance with HIPAA and to ensure that medical specialists or professionals who work with the Docus (the “Medical Expert”) do so as well.”
Privacy Policy, dated March 26, 2024; downloaded from the company’s site
Health data handled differently? Users told to remove identifiers before upload. The privacy policy says identifiable health information must not be uploaded and that you should remove your name, address and similar details first. It has no separate section for sensitive data, and no US state health-data notice was found.
“individually identifiable health information shall not be uploaded, shared and/or provided to the Docus and/or the Medical Expert, any and all direct identifiers of the User, such as name, address, or any other details that could identify an individual, should be removed”
Privacy Policy, dated March 26, 2024; downloaded from the company’s site
Whose AI reads the report? No statement found. We looked in the privacy policy, terms, four product pages and the trust center's Overview tab. No AI company or model is named. The four outside companies the trust center shows are cloud, monitoring and security providers.
Also worth knowing
- The trust center that the product page points to shows "HIPAA" as a bare badge label: no status word such as "compliant", no auditor, no date and no report to check.
- The trust center's only HIPAA sentence says practices follow "frameworks such as" HIPAA; the words "HIPAA compliant" do not appear on the saved page. “Our practices adhere to industry standards and frameworks such as SOC 2, HIPAA, and GDPR.” (Docus Trust Center, FAQ)
- No AI company is among the four outside processors the trust center overview shows (Amazon Web Services, Google Cloud Platform, Cloudflare and Vanta); the full list was not saved. “Amazon Web Services • Cloud provider” (Docus Trust Center, Subprocessors)
- The privacy policy allows sharing "the Information", its term for everything collected including medical information, with advertising and retargeting vendors. “We may share the Information with third-party vendors and service providers that help us with specialized services, including billing, payment processing, customer service, email deployment, business analytics, marketing (including but not limited to advertising, attribution, deep-linking, direct mail, mobile marketing, optimization and retargeting)” (Privacy Policy)
- The terms have you authorize the company to use your anonymized medical information to improve its website and services. “You authorize the Company to access, review, research, analyze, discuss, and copy your anonymized medical information. This information will be used solely for the purpose of enhancing the Website and providing improved Services.” (Terms of Use)
Limits of what we checked
- The privacy policy and terms are both dated 26 March 2024. The product pages show no date.
- Only the trust center's Overview tab was read, from a PDF saved on 2026-11-10. Its full Controls, Subprocessors and FAQ pages were not saved, so the lists shown may be incomplete. The trust center prints no date.
- The website pages were downloaded on 2026-11-09 and not re-checked on 2026-11-10.
- Docus also sells to labs; its home page is aimed at businesses. No page read says whether consumer and business customers are treated differently.
- The AI Doctor page makes a further claim ("fully compliant with HIPAA and GDPR standards"); the claim shown here is the one on the lab upload page.
- The privacy policy refers to a HIPAA notice that was not found linked on the pages read. No app store listing was found, so there is no store privacy label.
Kantesti AI Blood Test Analyzer (Kantesti Ltd)
What we looked at: Consumer web service, free and paid plans, as described on five pages saved on 2026-11-10
HIPAA claim in its marketing (a sentence; Home page, older "More questions" FAQ, answer to "Is my medical data secure when using the AI Blood Test Analyzer?"): “We implement HIPAA-compliant infrastructure, GDPR compliance protocols, and 256-bit end-to-end encryption for all data transfers and storage.”
Used to train AI? Terms: anonymized data may improve AI. The privacy policy says nothing about training AI on uploads. The Terms say anonymized, combined data may be used to improve the AI. No way to opt out was found. A page said to promise no training was not saved.
“Anonymized aggregate data may be used for AI improvement”
Terms of Use, section 9.3, dated March 17, 2026; read from an image of the page
How long is it kept? Free: up to 90 days; paid: while subscribed. The privacy policy's table says free-plan uploads are kept up to 90 days, and paid-plan uploads while you subscribe, then about 90 days more. You can ask for deletion by email or in account settings. The Terms say something different.
“Up to 90 days”
Privacy Policy, section 10 retention table, row "Uploaded health data (free tier)", dated March 13, 2026; saved from a browser as a PDF
Can people read it? No statement found. We looked in the saved privacy policy, Terms, home page and two Trust Center pages. The policy lists access controls for staff but does not say whether or when staff may look at an uploaded report.
Used for ads or sold? Says health data is not sold. The privacy policy says personal and health data is not sold. It lists sharing with a hosting company, a payment processor and an email/SMS provider. No statement about using health data for advertising was found. The home page disagrees on sharing.
“We do not sell your personal or health data.”
Privacy Policy, section 7, dated March 13, 2026; saved from a browser as a PDF
What it says about HIPAA Aims to comply where applicable; role unstated. The privacy policy says only that the company aims to comply with HIPAA, the US health privacy law, where applicable. No saved page says the law binds the company, and none offers consumers a HIPAA contract (a business associate agreement).
“We aim to comply with GDPR, HIPAA (where applicable), and other laws.”
Privacy Policy, section 1.3, dated March 13, 2026; read from an image of the page
Health data handled differently? Classed as sensitive; processed on consent. The privacy policy classes uploaded lab results as sensitive health data and says it processes them with your consent. A security page asks users to cover their own name and ID numbers before uploading. No US state health-data notice was found.
“Consent: process sensitive health data; optional communications; cookies/analytics where required.”
Privacy Policy, section 5, dated March 13, 2026; read from an image of the page
Whose AI reads the report? Described as proprietary; no outside provider named. The Terms call the AI model the company's own. No outside AI company is named on any saved page, and the privacy policy's sharing table has no AI provider row. The pages disagree about where the service is hosted.
“Proprietary multilingual neural network for blood-test report interpretation”
Terms of Use, section 2.1, dated March 17, 2026; saved from a browser as a PDF
Privacy label on the App Store: Data Linked to You: Health & Fitness, Contact Info, User Content, Identifiers, Diagnostics.
Also worth knowing
- The Terms say all data is purged straight after processing, while the privacy policy says uploads are kept for up to 90 days or for as long as you subscribe. “All data is immediately purged after processing. No patient data is stored permanently.” (Terms of Use, section 9.1)
- An older FAQ on the home page says blood test data is never shared with third parties, while the privacy policy lists a hosting company, a payment processor and an email/SMS provider that receive data. “Your blood test data is never shared with third parties, and you have complete control over your information.” (Home page, "More questions" FAQ)
- HIPAA is claimed at three strengths: the home page FAQ says "HIPAA-compliant infrastructure", the Terms list HIPAA under "Compliance Certifications", the privacy policy only "aims to comply", and the two Trust Center pages saved do not mention HIPAA at all. “Our Service maintains compliance with:” (Terms of Use, section 9.2 "Compliance Certifications")
- The pages disagree about hosting: the privacy policy and security page name PythonAnywhere, while the Terms say the service runs on Microsoft Azure. “Our Service leverages Azure's infrastructure, which provides:” (Terms of Use, section 5.3)
- The home page carries a "Made in Germany" badge, but every page's footer says the company is headquartered in London, and no German company or address appears on the saved pages. “Made in Germany since 2019” (Home page badge)
Limits of what we checked
- All five pages were read from PDFs saved from a browser on 2026-11-10. The PDFs print no web addresses; the URLs given were matched by page title from an earlier download.
- The privacy policy and Terms PDFs have a broken text layer, so most quotes from them were read by eye from the page image. Long lines are cut off at the page edge; quotes affected by that were not used.
- Two Trust Center pages (Privacy and Responsible AI) were not saved. Statements earlier attributed to them, including one that uploads are not used to train AI, could not be verified and are not used here.
- Kantesti's own pages contradict each other on retention, sharing, HIPAA, hosting and who the service is for. The cells give the privacy policy's position where it has one; the training and AI provider cells quote the Terms because the privacy policy is silent.
- The Terms read as if written for clinics and enterprise customers (they forbid direct-to-consumer medical advice), so their statements may not describe the consumer service; the Terms do not say.
- The App Store privacy label and seller name (an individual, not the company) were read in an earlier pass and not re-checked. The Google Play listing was not checked.
Levels (Levels Health, Inc.)
What we looked at: Free AI lab interpretation; the privacy policy describes lab uploads and AI features in general and does not name this feature
HIPAA claim in its marketing (a sentence; Home page, under the heading "Private, secure, and personalized"; a "HIPAA compliant" badge also appears there): “HIPAA compliant and SOC 2 compliant, with feedback that gets more personal as you connect more of your data.”
Used to train AI? Providers don't train; improvement opt-out by email. The privacy policy says the outside AI companies do not train their models on your data. Levels says it uses combined, de-identified data to improve its product; you can opt out of product-improvement and research use by email.
“We do not use your Levels data for advertising third-party products, and our AI service providers do not use your data to train its own models.”
Privacy Policy, "Levels Guide AI and similar features", dated June 5, 2026; downloaded from the company’s site
How long is it kept? Kept indefinitely unless deletion requested. The privacy policy says member data is kept while you are a member and indefinitely afterwards. You can ask by email for deletion, which it says happens within a month. It says it cannot delete data held by other companies.
“Levels retains member data on secure servers for the duration of its business relationship with the member, and for an indefinite period of time afterward.”
Privacy Policy, dated June 5, 2026; downloaded from the company’s site
Can people read it? Support staff and physicians can access. The privacy policy says lab data can be seen by authorized people who need it, such as independent physicians and support staff. It also says Levels may review transcripts of AI conversations with identifying details removed or replaced.
“Access to your lab data is limited to authorized personnel who need it to provide services, such as third-party / independent physicians for ordering and reviewing results, and support team members.”
Privacy Policy, dated June 5, 2026; downloaded from the company’s site
Used for ads or sold? Not sold; no third-party product advertising. The privacy policy says data is not sold and is not used to advertise other companies' products. It says data is shared with a published list of outside processors and that it tries to avoid advertising trackers.
“We don’t sell (or sell access to) your data to third parties (for advertising, or for other purposes).”
Privacy Policy, dated June 5, 2026; downloaded from the company’s site
What it says about HIPAA Says it is not a Covered Entity. The privacy policy says Levels is not a covered entity under HIPAA, the US health privacy law. It says it may act as a business associate only for clinician notes and telemedicine features. No HIPAA contract is offered to consumers.
“Levels does not provide medical care and is not a Covered Entity under HIPAA (US Health Insurance Portability and Accountability Act).”
Privacy Policy, "HIPAA (US only)", dated June 5, 2026; downloaded from the company’s site
Health data handled differently? Direct identifiers removed before AI sharing. The privacy policy says name, email, phone and address are removed before data goes to AI companies. It does not say whether names printed inside an uploaded lab report are removed. No US state health-data notice was found.
“Before sharing, we remove direct identifiers such as your name, email address, phone number, and street address, and instead associate your information with an internal user ID.”
Privacy Policy, dated June 5, 2026; downloaded from the company’s site
Whose AI reads the report? OpenAI, Anthropic, Google and xAI. The privacy policy names four AI companies that receive prompts and selected account data, including lab results. It does not say which of them processes the free lab interpretation. A fifth company, LangFuse, stores logs of prompts and responses.
“with our AI service providers (currently OpenAI, Anthropic, Google, and xAI) to provide those features”
Privacy Policy, "Levels Guide AI and similar features", dated June 5, 2026; downloaded from the company’s site
Also worth knowing
- The home page says "HIPAA compliant" without qualification, while the privacy policy says Levels is not a covered entity and ties its HIPAA role to clinician-note and telemedicine features, not to lab uploads. “Levels may act as a “business associate” to independent healthcare providers (HIPAA covered entities) under one or more Business Associate Agreements (BAAs).” (Privacy Policy, "HIPAA (US only)")
- The policy says Levels cannot delete copies of your data held by other companies. “We cannot delete your data that may be stored with any third parties.” (Privacy Policy)
- The AI companies may keep what they receive for a "limited period" whose length is not given. “Our AI service providers use this information only to provide services to Levels, do not use your Levels data to train their models, and may retain the data they receive for monitoring, abuse detection, debugging, and reliability purposes for a limited period.” (Privacy Policy, "Levels Guide AI and similar features")
- Levels says its staff may read transcripts of AI conversations after identifying details are removed or replaced. “We may review de-identified or pseudonymized transcripts of these interactions to troubleshoot issues, improve prompts, and monitor for abuse.” (Privacy Policy, "Levels Guide AI and similar features")
- On advertising trackers the policy promises effort, not a ban. “We strive to avoid using SDKs, code, or trackers provided by advertising networks that may track or retarget you for commercial purposes.” (Privacy Policy)
Limits of what we checked
- The privacy policy does not name the free AI lab interpretation. We treated its wording about lab uploads and AI features as covering it, but the pages do not say so.
- The privacy policy is a Notion page that needs JavaScript; its text was read from the page's public data, not from the rendered page. Its dates (effective 2022-05-13, last updated 2026-06-05) were read as data fields, and the page's own record shows a later edit on 2026-08-21.
- The policy does not say which of the four AI companies handles the free lab interpretation, or how long they keep data.
- The deletion promise covers "identifiable" data; nothing is said about deleting de-identified or combined data.
- The upload flow itself, the cookie policy and the mobile app's App Store listing were not checked, so no store privacy label is recorded.
NoctorAI / Visualizer (Sage Bio)
What we looked at: The consumer Visualizer and NoctorAI at sagebio.com
HIPAA claim in its marketing (a sentence; NoctorAI product page): “Our HIPAA-compliant AI, known as "NoctorAI," helps users understand their lab results without a medical background or the oversight of a medical provider.”
Used to train AI? No statement found. No page says whether uploaded reports are used to train AI models, and no opt-out was found. We read the privacy policy and terms. The terms say data sent to the file reader is used solely for visualization and insights; the policy allows improving services.
How long is it kept? Up to 12 months after account goes idle. The policy says health data is kept as long as you have an account, and no longer than twelve months after the account goes idle. Deleting on request removes data from active databases, though the policy says some information may be kept.
“No purpose in this notice will require us keeping your personal information for longer than twelve (12) months past the start of the idle period of the user's account.”
Privacy Policy; downloaded from the company’s site
Can people read it? No statement found. No page says whether Sage Bio staff or contractors can view reports uploaded by consumers. We read the privacy policy, terms of use, the NoctorAI, Visualizer, About and For Clinics pages, and the home page.
Used for ads or sold? Says not sold; advertising trackers permitted. The policy says no personal information was sold or shared in the past twelve months. It also says outside companies may use tracking on the service for advertising, and that data is shared with service providers. The terms allow unrestricted disclosure of de-identified data.
“We have not sold or shared any personal information to third parties for a business or commercial purpose in the preceding twelve (12) months.”
Privacy Policy; downloaded from the company’s site
What it says about HIPAA No statement found. Consumer pages and the privacy policy call the AI and platform HIPAA-compliant, but no page says how the law applies to consumer use or that Sage Bio is a covered entity or business associate. A business associate agreement is offered to clinics only.
Health data handled differently? Health data processed on explicit consent. The policy has a sensitive-information section that names health data from uploaded lab results and says it is processed on your explicit consent. No Washington-style consumer health data notice was found, and Washington is not among the states the policy lists.
“Specifically, for the processing of sensitive health data through LabLynx and NoctorAI, we rely on your explicit consent as the legal basis.”
Privacy Policy; downloaded from the company’s site
Whose AI reads the report? No statement found. NoctorAI is described as Sage Bio's own medically fine-tuned language model. No underlying model or outside AI company is named on the privacy policy, terms, product pages or cookies policy, and nothing is said about a provider's use of data.
Also worth knowing
- A business associate agreement, the contract HIPAA requires when a company handles patient data for a clinic, is offered on the page for clinics; no such wording exists for consumers. “Business Associate Agreement provided” (For Clinics page)
- The privacy policy repeats the HIPAA-compliant label for the AI but handles consumer health data on the basis of consent and state privacy laws, without saying what the label means. “This extracted data is then securely fed into NoctorAI, our medically fine-tuned and HIPAA compliant LLM, to generate interpretations and provide lifestyle-related advice based on your lab values.” (Privacy Policy)
- The policy permits outside advertising trackers on the service, and no statement was found that health data is excluded from them. “We also permit third parties and service providers to use online tracking technologies on our Services for analytics and advertising, including to help manage and display advertisements, to tailor advertisements to your interests, or to send abandoned shopping cart reminders (depending on your communication preferences).” (Privacy Policy)
- The terms let the company turn personal information into de-identified or combined data and disclose that data without restriction. “Convert or combine some personal information into de-identified or aggregated data that does not disclose any of the personal information of any individual Client or Visitor, and may disclose any such de-identified or aggregated data without restriction;” (Terms of Use)
- The privacy policy names a Kansas company, Sage Bio, Inc., while the terms of use name Sage Bio Solutions Corp and choose the law of a Canadian province. “shall be governed and construed according to the laws of the Province of Alberta” (Terms of Use)
Limits of what we checked
- The privacy policy and terms carry no date, although the policy says changes will be marked by a revised date at the top.
- The privacy policy and the terms of use name different companies and jurisdictions; which one operates the consumer product is not stated.
- The retention limit depends on an "idle period" that the policy does not define.
- The app's sign-up page could not be read, so any consent text shown there is not covered.
- No App Store listing for Sage Bio was found, so no store privacy label is recorded.
- Sage Bio also sells to clinics; the business associate agreement and other clinic terms were not read beyond the For Clinics page.
SelfDecode (Decodify Me LLC)
What we looked at: Lab Analyzer and the "AI Hyperintelligence" assistant, consumer accounts
HIPAA claim in its marketing (a sentence; Lab Analyzer subscription FAQ in the help centre): “Your data is protected with HIPAA and GDPR compliance”
Used to train AI? Says no AI training on your information. The policy says your information is not used to train AI models. It also says conversations, which include uploaded files, are used to review and improve the AI feature. The only opt-out is withdrawing AI consent, which switches the feature off.
“We do not use your information to train or fine tune AI models, and we do not permit anyone else to do so.”
Privacy Policy; downloaded from the company’s site
How long is it kept? While account open; in-app deletion within one day. The policy says information is kept while your account is open. Deleting in the app removes it within one day; deletion through customer support takes up to 30 days. Data used in completed research or kept for legal reasons may not be erased.
“Individual conversations you delete, and accounts you delete from within the app or website (together with all associated information, including Conversation Information), are deleted from our systems within one day.”
Privacy Policy; downloaded from the company’s site
Can people read it? Authorised staff may review AI conversations. The policy says authorised SelfDecode staff may review AI conversations, which include uploaded files, to improve accuracy and safety. A help article also describes a service in which the company's team uploads your lab results for you.
“authorized SelfDecode personnel may review conversations for this purpose”
Privacy Policy; downloaded from the company’s site
Used for ads or sold? Sensitive information not sold or shared for ads. The policy says sensitive information is not sold or shared for advertising that follows you across sites. It also says lab information goes to outside processing companies and that combined, de-identified data may be shared with third parties.
“We do not sell your Sensitive Information, and we do not share it for cross context behavioral advertising.”
Privacy Policy; downloaded from the company’s site
What it says about HIPAA No statement found. Marketing and help pages claim HIPAA compliance, but the privacy policy, terms and platform consent never mention HIPAA. No page says how the law applies to SelfDecode, that it is a covered entity or business associate, or that consumers are offered an agreement.
Health data handled differently? AI conversations treated as health information. The policy treats everything in an AI conversation as health information and counts lab results as sensitive information. It says anonymised, combined lab data may be used for research, product development and promotion. No Washington-style consumer health data notice was found.
“Because Conversation Information typically concerns your health, we treat all of it as health information, whether or not a particular message does.”
Privacy Policy; downloaded from the company’s site
Whose AI reads the report? Unnamed outside model, run in SelfDecode's cloud. The policy says the AI model was developed by another company, which it does not name, and runs in a cloud environment SelfDecode controls. It says the model's developer cannot access your information.
“The AI model that powers Hyperintelligence was developed by a third party but runs within a cloud environment that SelfDecode controls, operated by the same third party cloud hosting provider described under Data Storage and Security.”
Privacy Policy; downloaded from the company’s site
Privacy label on the App Store: Purchase history, coarse location, user ID and product interaction listed for the developer's advertising or marketing; no health data category listed.
Also worth knowing
- The home page promises never to share your data, while the privacy policy describes sharing lab information with outside processing companies and sharing personal information with companies under common ownership. “We will never share your data” (Home page)
- The home page says the company follows HIPAA policies, but the word HIPAA does not appear anywhere in the privacy policy or the terms of service. “We follow HIPAA and GDPR policies” (Home page)
- A help article calls the AI "our own", while the privacy policy says the model was developed by a third party; the two fit together only if "our own" means run on SelfDecode's own systems. “The AI in V4 is our own, and your data never leaves our platform — we don't send it to any outside AI provider.” (Help centre, SelfDecode V4)
- A help article says deletion takes up to 24 hours, which differs from the policy's 30 days for deletion requested through customer support. “Deletion takes up to 24 hours to finalize.” (Help centre, data security)
- The policy allows lab results, once anonymised and combined with other users' results, to be used for promotion as well as research. “We may combine your Laboratory Information with that of other users in an anonymized and aggregated form (“Aggregated Laboratory Information”) for research, product development, and promotional purposes.” (Privacy Policy)
Limits of what we checked
- The privacy policy prints no effective or last-updated date; the page's hidden metadata says it was last modified on 25 September 2026.
- The no-training sentence sits in the policy's section on AI-powered features (the Hyperintelligence assistant); the older PDF upload path ("Input Labs") is not separately addressed.
- The policy does not name the AI model's developer or the cloud provider; a help article names Amazon Web Services only as where genetic data is stored.
- SelfDecode also sells a product for practitioners (PromicsEdge); its help articles and any HIPAA agreement terms for practitioners were not read.
- An optional connection to your own ChatGPT or Claude account exists; how data is handled on that path was not checked.
- The App Store label belongs to an app whose last listed version is 1.1, dated 07/10/2022; the Google Play listing was not checked.
- The company's country and address are not stated on the pages read.
SiPhox Health (upload analyzer, "Sai") (SiPhox, Inc.)
What we looked at: The upload-your-results analyzer at siphoxhealth.com/upload-results, United States
Used to train AI? Policy allows model training; no opt-out found. The privacy policy says your information, mainly in anonymised or combined form, may be used to develop and train models. No setting to turn off AI training was found. The upload page says uploads are used solely to generate insights.
“We may use your information—primarily Aggregated Data (defined below)—for the development, training, and validation of predictive models, health support tools, insights, and other algorithmic systems aimed at improving diagnostic accuracy, product quality, and personalization of services.”
Privacy Policy, dated November 7, 2026; downloaded from the company’s site
How long is it kept? No time limit stated; deletion offered. The upload page says you can delete your personal information at any time. No page says how long uploads are kept. The terms say the company may keep server copies of deleted user content, which may or may not cover uploaded reports.
“Delete your personal information from our servers at any time. We respect your right to data ownership and provide easy-to-use tools for data management.”
Upload page; downloaded from the company’s site
Can people read it? No statement found. No page says whether staff can view uploaded reports. We read the privacy policy, upload page, FAQ and parts of the terms. The policy says information is used for customer support and may be disclosed to contractors and licensed healthcare professionals.
Used for ads or sold? Says not sold; anonymised data may be sold. The privacy policy says personal and health information is not sold. The same policy says anonymised, combined data may be sold for medical research and used for marketing insights and targeting, and that the website uses behavioural tracking.
“We do not sell your personal or health information.”
Privacy Policy, dated November 7, 2026; downloaded from the company’s site
What it says about HIPAA No statement found. The word HIPAA does not appear on the upload page, home page, FAQ, privacy policy or Washington policy. The terms mention it only for requesting your records from labs and providers. No page says the company itself is bound by HIPAA.
Health data handled differently? State health-data notice; separate Washington policy. The privacy policy has a notice for residents of states with health-data laws, and there is a separate policy for Washington residents. Both say data that has been anonymised can be used without the limits that apply to personal information.
“If you are a resident of certain U.S. states, including Washington, Connecticut and Nevada, you may be entitled to additional rights and disclosures, for your health information/health data, depending on the law of your state.”
Privacy Policy, dated November 7, 2026; downloaded from the company’s site
Whose AI reads the report? OpenAI and ChatGPT named as examples. The privacy policy lists using anonymised data with AI programs "such as ChatGPT", and the terms say providers "such as OpenAI" may power the Sai chatbot. The company whose AI reads the uploaded file is not named.
“use with artificial intelligence models and programs, such as ChatGPT, to provide insights for a consumer’s test results”
Privacy Policy, dated November 7, 2026; downloaded from the company’s site
Also worth knowing
- The upload page says uploaded results are used solely to generate insights and an action plan, which the privacy policy does not match. “We take your privacy seriously. When you upload third-party test results, your data is used solely to generate insights and an action plan. We do not share, sell, or use your information for any other purpose.” (Upload page FAQ, "How is my data handled?")
- The privacy policy, by contrast, allows information to be used for AI development, including personal information where anonymised data is not possible. “Personal information used for AI development and improvement is Aggregated Data whenever possible. If personal information is used, it is processed in accordance with applicable privacy laws and only for legitimate purposes as outlined in this Privacy Policy.” (Privacy Policy)
- The upload page banner says "Never sold", while the privacy policy lists selling anonymised, combined data among its uses. “selling Aggregated Data for medical research/bio tech research purposes.” (Privacy Policy)
- The terms say the company may keep server copies of content you have deleted; it is not certain that an uploaded lab report counts as this kind of content. “Notwithstanding the foregoing, you understand and agree that SiPhox may retain, but not display, distribute, or perform, server copies of User Content that have been removed or deleted;” (Terms of Service, section 2.3)
- The terms say consent to the use of your data is withdrawn by sending a written notice to a postal address, and that without consent you may lose access to the service. “I understand that I have the right to revoke this consent, in writing, at any time, by sending written notification to: SiPhox Health, 111 Terrace Hall Ave, Burlington, MA 01803.” (Terms of Service, section 17.0)
Limits of what we checked
- The privacy policy and terms are written for all of SiPhox's services, not for the upload feature alone; most quoted policy passages are not specifically about uploaded reports.
- The privacy policy and terms were both changed on 7 November 2026, three days before we checked.
- The logged-in privacy settings page was not read, so we do not know whether it contains a switch for AI training.
- The terms of service were searched by keyword and read in the relevant sections, not from end to end.
- The passages naming OpenAI and ChatGPT are about the chatbot and about anonymised data; no page says whether names and birth dates are removed from an uploaded report before an outside AI service processes it.
- The policy refers to a separate notice about processing health information that we could not locate.
- The SiPhox mobile app's store listing was not checked, so no store privacy label is recorded.
SnapLabs (Zujunai (Lithuania))
What we looked at: The SnapLabs app and snaplabs.app
Used to train AI? Says not used to train AI models. The policy says the company has opted out of training with its AI provider, Google, so medical data is not used to train the AI models. This is the company's own setting, not a choice offered to users.
“Zero Training: We have opted out of data training. Your medical data is not used to train the AI models.”
Privacy Policy, dated September 23, 2026; downloaded from the company’s site
How long is it kept? On device by default; deletable in app. The policy says health data stays on your device by default. If you turn on Data Sync, it is stored in the company's database with no stated time limit. The policy says deleting your account permanently wipes data from your device and its servers.
“Default Local Storage: By default, all of your health data (documents, metrics, records, and insights) are stored locally on your device and are not stored on our servers.”
Privacy Policy, dated September 23, 2026; downloaded from the company’s site
Can people read it? Says it does not view raw photos. The policy says original images stay on your device and the company does not view your raw photos. No statement was found on whether staff can see extracted values or documents once Data Sync is turned on.
“Original images are stored on your device. We do not view your raw photos.”
Privacy Policy, dated September 23, 2026; downloaded from the company’s site
Used for ads or sold? App Store description: no ads, no selling. The App Store description says there are no ads and data is never sold. The privacy policy has no sentence on sale or advertising; it lists the outside companies used: Google for AI, Supabase for storage and sign-in, and RevenueCat for subscriptions.
“No Ads, No Selling: We never sell your data to third parties.”
App Store description; read from the App Store listing
What it says about HIPAA No statement found. HIPAA is not mentioned on the privacy policy, terms, AI transparency page, instructions for use, home page or App Store listing. The company makes no HIPAA claim and describes itself under European Union and Lithuanian law.
Health data handled differently? Consent required; whole document sent to AI. The policy names health data as its own category and the app requires your consent before AI extraction. The policy says the whole document, including personal details printed on it, goes to the AI model. No Washington-style consumer health data notice was found.
“The whole document is sent for extraction, so any personal details printed on it (such as your name) are processed by the AI model.”
Privacy Policy, dated September 23, 2026; downloaded from the company’s site
Whose AI reads the report? Google (Gemini 3.1 Flash-Lite). The policy says images of your documents are sent to Google and processed by its Gemini 3.1 Flash-Lite model. The AI transparency page names Google Ireland Ltd. and says the company has opted out of training.
“images of your documents are transmitted securely (encrypted) to our third-party AI provider (Google), where they are processed by the Gemini 3.1 Flash-Lite model”
Privacy Policy, dated September 23, 2026; downloaded from the company’s site
Privacy label on the App Store: Data Not Collected. The App Store label says no data is collected, while SnapLabs' own privacy policy says it collects email address, name, usage and crash metrics, and health data when Data Sync is on.
Also worth knowing
- The App Store privacy label says no data is collected, which does not agree with the company's own privacy policy. “Data Not Collected” (App Store privacy label)
- Turning on Data Sync moves health data from your device to the company's database, and no time limit is stated for how long it stays there. “If enabled, your data is synced to our secure database (Supabase).” (Privacy Policy)
- The AI transparency page says Google processes the data through what the company calls an Enterprise API with training turned off; no statement was found on how long Google keeps what is sent. “Your data is processed via their Enterprise API. We have opted out of data training, meaning your health data is NOT used to train their models.” (AI Transparency & Technologies)
- The home page footer names "SnapLabs Inc.", but the privacy policy and App Store listing name Lithuanian businesses and no "Inc." appears in the policy or terms. “© 2026 SnapLabs Inc.” (Home page footer)
Limits of what we checked
- The company's pages give three slightly different names: "Zujunai" in the privacy policy, "ZUJUNU MALUNAS, MB" as the App Store seller, and "SnapLabs Inc." in the home page footer.
- The no-ads and no-selling statement appears only in the App Store description, not in the privacy policy.
- Account details (email address, subscription status) are held by outside services whether or not Data Sync is on.
- "Enterprise API" is the company's wording; the pages do not say which Google product is meant, and no data processing agreement with Google is mentioned.
- The Google Play listing and the consent screens inside the app were not checked.
- Pages were downloaded on 9 November 2026.
Wizey (Wizey LLC)
What we looked at: Consumer service at wizey.one and its iPhone app
HIPAA claim in its marketing (a sentence; Home page, privacy section): “The service operates in line with HIPAA & GDPR data protection standards.”
Used to train AI? No statement found. No page says whether uploaded reports are used to train AI models, and no opt-out was found. We searched the privacy policy, terms, both data processing documents and the home page. The policy lists "improving our Services" as a use without explaining it.
How long is it kept? As long as necessary; deletion offered. The privacy policy gives no fixed period, saying data is kept only as long as necessary. The home page says you can delete your account and data at any time. No page describes what deletion covers, such as backups or copies at AI providers.
“We retain data only as long as necessary:”
Privacy Policy, dated April 19, 2026; downloaded from the company’s site
Can people read it? No statement found. No page says whether Wizey staff or contractors can view uploaded reports. We searched the consumer and legal pages. The home page only says documents are never opened to other users, and the policy lists access controls.
Used for ads or sold? Says never sold or used for advertising. The home page says data is never used for advertising and never sold. The privacy policy itself has no such sentence; it allows sharing with service providers and partners, affiliated companies, and in business transfers, without naming any of them.
“Data is encrypted in transit and at rest, never used for advertising, and never sold to third parties.”
Home page; downloaded from the company’s site
What it says about HIPAA No statement found. Wizey's privacy policy and marketing pages claim HIPAA compliance, but no page says how the law applies to Wizey. Nothing says it is one of the organisations HIPAA binds (a covered entity or business associate), and no business associate agreement is offered.
Health data handled differently? Says identifiers removed before AI analysis. A separate AI notice says names, birth dates and ID numbers are removed before AI analysis. The privacy policy says sensitive health data is processed only with your explicit consent. No Washington-style consumer health data notice was found.
“Before AI analysis, all personally identifiable information (such as names, dates of birth, and identification numbers) is automatically removed.”
AI Data Processing, dated April 19, 2026; downloaded from the company’s site
Whose AI reads the report? No statement found. The pages refer to third-party AI services without naming any company. We searched the legal documents, home page and business pages for the names of the major AI providers and found none. Nothing is said about how those providers keep or use data.
Privacy label on the App Store: Data Not Collected. The App Store label says the developer collects no data from the app, while Wizey's privacy policy says it collects name, email address, health-related information, IP address, device information and usage data.
Also worth knowing
- The App Store privacy label says no data is collected, which does not agree with Wizey's own privacy policy. “The developer does not collect any data from this app.” (App Store privacy label)
- The strength of the HIPAA claim changes from page to page: "in line with" on the home page, "aligned" on the contact page, and "complies with" on the page for business partners. “The service complies with HIPAA and GDPR requirements.” (Security page for business partners)
- The legal index says the privacy policy covers your rights under HIPAA, but no HIPAA-specific rights section was found in the policy. “What data Wizey collects, the legal basis, how it is protected, and your rights under HIPAA, GDPR and CCPA.” (Legal index)
- The AI notice describes only lab values being sent to outside AI services, although the product also accepts free-text documents such as discharge summaries and prescriptions; it does not say how those are anonymised. “Only anonymized medical data (laboratory values, test names, and reference ranges) is sent to third-party AI services for interpretation.” (AI Data Processing)
- The website names Wizey LLC of Wyoming as the company, while the App Store listing names an individual as the seller and does not mention Wizey LLC.
Limits of what we checked
- The promises that data is never sold and never used for advertising appear on the home page and in the App Store description, not in the privacy policy.
- Wizey also offers its service to business partners; its Data Processing document is written partly as a business agreement, so it is unclear which of its retention terms apply to consumer accounts.
- The home page also shows a "Privacy HIPAA / GDPR" badge on a feature card; the HIPAA claim recorded here is the home page sentence, and stronger wording appears on the page for business partners.
- The signed-in web app and any consent text shown inside the app were not read.
- The Google Play listing and its data safety section were not checked.
- Pages were downloaded on 9 November 2026.
A HIPAA claim rarely comes with an explanation
The apps that advertise HIPAA are BloodGPT, Docus, Kantesti AI Blood Test Analyzer, Levels, NoctorAI / Visualizer, SelfDecode, Wizey. HIPAA binds two kinds of organization: “covered entities” such as providers and insurers, and the “business associates” that handle records for them under a signed agreement. We looked for any sentence saying which of those the company is when a consumer uploads a report.
2 of the 7 address the question, and none of those says the law covers the consumer:
- Docus: “Although our Services may not include the specific transactions to which HIPAA may apply, we have nonetheless decided to operate in compliance with HIPAA and to ensure that medical specialists or professionals who work with the Docus (the “Medical Expert”) do so as well.”
- Levels: “Levels does not provide medical care and is not a Covered Entity under HIPAA (US Health Insurance Portability and Accountability Act).”
The rest make the claim and leave it there. BloodGPT shows it only as a badge image, with no sentence behind it. Where a business associate agreement is mentioned at all, it is offered to clinics, not to individuals. Supporting material is thin even where it exists: the trust page that Docus’s marketing points to shows the word “HIPAA” as a label, with no auditor, date, or report.
Store privacy labels that disagree with the policy
Apple asks developers to declare what their app collects, and shows the answer on the App Store page. The declaration is the developer’s own. For 3 apps the label reads “Data Not Collected” while the company’s privacy policy lists what it collects:
- BloodTracker AI: The company's privacy policy lists collection of name, date of birth, health data, device information and IP address, and sharing with OpenAI, Google Firebase and AppsFlyer.
- SnapLabs: The App Store label says no data is collected, while SnapLabs' own privacy policy says it collects email address, name, usage and crash metrics, and health data when Data Sync is on.
- Wizey: The App Store label says the developer collects no data from the app, while Wizey's privacy policy says it collects name, email address, health-related information, IP address, device information and usage data.
We saw the same pattern on Google Play among apps that did not meet our rules for inclusion: several whose listings describe uploading reports for AI analysis declare “No data collected.” Google’s form counts only data sent off the device, so those declarations are wrong only if the analysis happens on a server, and the listings do not say.
When a company’s pages disagree with each other
A privacy policy is one page among several, and the marketing page is the one people read. Several apps say one thing on the home page and another in the policy. The details are under each name above; the kinds of disagreement we found were:
- a home page saying data stays on your device, and a policy saying reports are sent to an outside AI company;
- a home page saying data is “never shared,” and a policy listing the companies it is shared with;
- terms saying data is purged immediately after processing, and a policy on the same site saying uploads are kept for up to 90 days;
- “never sold,” beside a policy allowing the sale of anonymized data.
We are not in a position to say which page is right. A reader cannot tell either, which is the problem.
Who actually runs the AI
7 of the 12 apps give some statement about whose AI model processes the report, and fewer name the company outright. This matters because the app’s promises about training and retention bind the app. What the AI company behind it may keep is governed by a contract you cannot see.
What to do with this
If you want an AI’s help with a lab report and would prefer the report not sit on someone’s server:
- Run a model on your own computer. Nothing leaves the machine. The walkthrough is here.
- Remove what identifies you first. A lab report carries your name, date of birth, and often an address and account numbers. The values are what the AI needs.
- Turn off the training setting, and use a temporary chat where one exists. Read the product’s entry above for what that setting does not cover.
- Delete the file, not only the chat, and check for saved memories.
- Treat “HIPAA compliant” on a consumer app as a statement about security at most. Look for what the policy says about training, retention, and who else receives the data.
HealthViewer, which we make, is a lab results tracker that runs on your own computer and keeps your results there, so we have an interest in this subject. That is why this page quotes and links everything, and why we would rather you check it than take our word. If your aim is to follow results over the years and not to ask an AI about one report, how to track lab results over time covers the options.
How we did this
The questions. For each product: whether chats and uploads are used to train AI and whether you can turn that off; how long they are kept and what deleting does; whether people at the company can read them; whether they are used for advertising or sold; what the company says about HIPAA for consumer use; and whether health information is handled differently. For the lab-result apps we also asked which company’s AI model processes the report.
The pages. We read each company’s privacy policy, terms, and the help or product pages that address these questions, on November 9, 2026 and November 10, 2026. Most were downloaded directly. Some sites refuse automated requests, and for those a person opened each page in a browser and saved it as a PDF. Meta’s pages were read from Internet Archive copies. Each quotation above says which.
The quotations. Every quotation was matched, character for character, against the text of the page it came from, and the page that builds this article refuses to publish a quotation that does not match our saved copy. Passages that were cut off, reconstructed, or only summarized were left out. Where a page prints a date, we give it.
“No statement found” means we looked in the pages listed for that product and found nothing on the question. It does not mean the company does or does not do something.
Choosing the lab-result apps. We ran nine searches a patient might run (five on the web, two in the App Store, two on Google Play) and recorded the first ten products from each. A product was included if it is sold to consumers, accepts an uploaded lab report and interprets it with AI, is offered in English to people in the United States, links a privacy policy, and either appeared in two or more of the nine searches or advertises HIPAA compliance on the pages we read. Two of the twelve fit those rules loosely, and their entries say how: one app’s privacy policy link did not lead to a readable policy, and one company’s pages do not say whether it serves the United States. Seven further apps found only on Google Play were screened and did not meet the rules.
The chatbots are seven widely used general-purpose assistants that accept an uploaded file or image.
What this does not show
- Practice. A policy is a statement. We did not inspect any company’s systems, and a company may do less, or more, than its policy allows.
- The future. These pages change often; several had been revised within days of our reading. The date on this page is the date of the answers.
- Other countries and other plans. Everything here is the US consumer product. People in the EU and UK have rights these pages do not describe, and business plans differ.
- Everything a company has published. We read the pages that bear on these questions, not every page. Each product’s entry lists what we did not reach.
- A ranking. The products differ in what they are for, and we have not scored them.
One of the apps, SnapLabs, also appears in our own comparison of lab result trackers without a subscription; it is here because it met the rules above.
Corrections
If you work for one of these companies and something here is wrong or out of date, write to support@healthviewer.net with the page that shows it. We will correct the entry and note the change.
Citing this page
HealthViewer, “What AI chatbots say they do with a lab report you upload,” answers as of November 10, 2026. The answers and quotations are also available as structured data.