Terms of use
These terms are between you and Polygraph, Inc., which publishes Polygraph. By using Polygraph you agree to them and to the privacy policy. Polygraph isn’t made or endorsed by OpenAI, Anthropic or any AI app it works with.
What Polygraph does
Polygraph helps one person ask another a few specific questions, answered from their own records. Polygraph drafts the questions, with any definitions, the evidence criteria each answer must meet and the form it takes; the requester approves them, and the person asked agrees to the whole request or not. After agreeing, the person asked’s own AI agent looks through their records privately and shows them the answers it proposes. Once they approve, it sends Polygraph an answer to every question, all together. Polygraph shows both people the answers and the date they were verified. An answer carries no evidence.
What a verified answer means
- A verified answer means the asked person’s AI agent reported that it established that answer, under the agreed questions and criteria, from records it examined. Polygraph doesn’t see those records and can’t independently authenticate them. Polygraph checks an answer’s form and whether it fits its question, not whether it is true. It is not a guarantee that the answer is correct, and it is not a cryptographic proof.
- No, zero and None are answers like any other. A request that isn’t answered simply has no answers. That is never a finding about what they would be, and declining, not answering or a request ending are not accusations.
- Each answer concerns its own question’s wording, definitions, criteria and dates. It isn’t refreshed or monitored afterwards.
- A request is open for 14 days from when it is first sent. After that nothing more can be agreed or answered on it, and answers already published stay as they are.
- AI can make mistakes at every step, from drafting questions to looking through records.
- Answers aren’t legal, financial, employment or other professional advice.
No automated decisions about people
Don’t use Polygraph or its answers as the only basis for a decision with legal or similarly significant effects on a person, such as hiring or firing, credit, housing, insurance, education or access to essential services. A person must weigh the answers, and any request left unanswered, alongside other relevant information, and the person concerned must have a fair chance to respond. Despite its name, Polygraph doesn’t detect lies, judge honesty or character, or score people.
Consent
- A request is sent only after the requester approves its exact questions. The person asked decides freely whether to agree to the whole request, suggest changes or decline, and declining is never treated as an answer. Their AI agent looks through their records only after they agree, and sends answers only after they approve them.
- Don’t use threats, deception or pressure to obtain agreement.
- Approvals and decisions made through your account, including by an AI agent you have connected and asked to act, are your decisions.
Restricted information
Don’t use Polygraph to request, submit or share:
- payment card details;
- health information;
- government identification numbers, such as social security, passport or driving licence numbers;
- passwords, API keys, sign-in codes or other secrets;
- other sensitive personal information, such as someone’s racial or ethnic origin, religion or beliefs, political views, trade union membership, sex life or sexual orientation, genetic or biometric data, or criminal record.
Where a fact depends on such information, ask for a narrower answer instead, such as whether someone holds a current licence rather than its number.
Acceptable use
- Follow the law, and the rules of the AI app you use Polygraph through.
- Ask only for information the person asked is entitled to share, and share only information you have the right to share. That includes other people’s personal information and anything you hold in confidence.
- Don’t use Polygraph to harass, stalk, monitor, discriminate against or profile anyone.
- Don’t impersonate anyone, or send an answer, or direct your AI agent to, that the records don’t establish under the agreed questions and criteria.
- Don’t try to get around Polygraph’s checks, including those on which AI apps can send answers, or disguise the app you are using.
- Don’t interfere with the service, probe it for weaknesses without permission, or use it at a volume meant to degrade it.
Your account
You’re responsible for what happens through your account and the AI apps you connect to it. Keep your sign-in secure, and tell us at support@usepolygraph.com if you think someone else has used it.
Your content
You keep whatever rights you have in what you submit. You allow Polygraph to store, process and share it as these terms and the privacy policy describe, to run the requests you take part in. Answers are shared with the two people in a request. Don’t present them as a certification or guarantee by Polygraph.
Availability
Polygraph is new and changing. Features may change or stop, and the service may be interrupted. Not every AI agent can send answers; each request says whether yours can. We may limit or suspend access to stop misuse or protect other people. You can stop using Polygraph at any time and ask us to delete your information; the privacy policy explains how we review those requests.
No warranties
Polygraph is provided as it is and as available. To the extent the law allows, we make no promise that it will be uninterrupted, error-free or fit for a particular purpose, or that any answer is correct. Nothing in these terms limits rights you have by law that can’t be limited.
Changes
We’ll post changes to these terms here with a new date at the top.
Contact
Write to support@usepolygraph.com.