All documents

Draft — under attorney review. Tovkin is for people who live in the US.

Tovkin AI Disclosure

Draft — requires attorney review. Not legal advice. Short version for the app (shown in onboarding, in Settings → About, and linked from every companion screen), plus a long version.

Short version (in-app, under the companion's name)

Tovkin is an AI. It reads what you type and upload and writes answers. It isn't a doctor, can't examine you, and can be wrong. It gives health information, not medical advice. Talk with a clinician about your results. In an emergency, call 911. If you're thinking about hurting yourself, call or text 988.

First message the companion sends in every new conversation:

Hi, I'm Tovkin, an AI health companion. I can explain things in plain English and help you prepare for your clinician, but I'm not a doctor and I can't see or monitor you. If this is an emergency, call 911.

If asked "are you a human?" or "is this a real doctor?" the companion answers truthfully every time: "No. I'm an AI." When a person at Tovkin writes to a member (for example about a request), the message is labeled "From the Tovkin team".

Long version

What the AI does

  • Answers general health questions in plain English.
  • Explains lab results and reports you upload, quoting your numbers and the lab's reference range.
  • Suggests questions to ask your clinician and prepares a brief for your visit.
  • Remembers what you told it earlier in your Tovkin account so you don't have to repeat yourself.
  • Helps draft requests (such as booking or records) for you to approve.

What it does not do

  • It does not diagnose, treat, prescribe, or tell you to start, stop or change medicines.
  • It does not examine you, see you, or watch your health in real time.
  • It does not replace your clinician, emergency services, or a crisis line.
  • It does not provide therapy or psychotherapy.
  • It does not act outside the app unless you approve the specific request.

Limits and risks

  • It can be wrong. AI can misread a photo, miss context, state something confidently that isn't true, or give an out-of-date answer. We reduce this with fixed safety rules and approved reference ranges and guidelines, and a clinician reviews our content, but errors can still happen.
  • It doesn't know everything about you. It only knows what you've given it.
  • Check anything important with a clinician. Especially symptoms, results outside the range, medicines and anything that worries you.
  • Emergencies bypass the AI. If you describe something that could be an emergency, Tovkin shows a fixed message telling you to call 911 or 988 and nothing else.
  • Studies have found that general-purpose AI chatbots often give problematic health advice. Tovkin is built differently, but is not immune to error. [CLINICAL: confirm the BMJ Open 2026 citation before using it publicly; see CLAIMS-AUDIT.]

How your data is used

Your messages and files are sent to an AI service provider to generate answers, under contract, and are not used to train AI models [CONFIRM]. See the Privacy Policy section 4. A person at Tovkin may read conversations to help with a request, investigate a safety issue or fix a problem. Tovkin doesn't use your health data for ads.

People

In the early-access period, requests such as bookings and record requests are handled by a person at Tovkin (currently the founder), who may use AI tools to prepare them. A person approves outbound actions only after you approve first. [CONFIRM wording against how legwork is built; do not say Tovkin "calls" if a human or an integration does it.]

Report a problem

Use "This was wrong or unsafe" under any answer, or email hello@tovkin.com. We treat these as safety reports and review them.