Acceptable use

Short version: use it for work, do not use it to hurt people, and do not try to break it.

Do not

Send anything you are not allowed to send to a third party. Your requests go to Anthropic and OpenAI — if you are under an obligation of confidence to someone else, that obligation travels with the text.

Use it to produce material that harms people: harassment, sexual content involving minors, instructions for weapons or attacks, or content designed to deceive someone into losing money.

Use it to impersonate a real person or organisation, or to produce fabricated records, reviews, or credentials.

Attempt to extract the models' underlying instructions, resell access to the models through this product, or use it to build a competing model.

Attempt to reach another customer's data, probe the service for vulnerabilities without asking us first, or automate signups.

The providers have their own rules

Anthropic and OpenAI apply their own policies to what you send, and either may decline a request. When that happens you will see it — we do not quietly substitute a different answer.

If you break this

We will suspend the account and tell you why. If it was a misunderstanding, say so and we will sort it out. If it was deliberate and serious, we will close the account and refund nothing.

Reporting a security problem

If you find a vulnerability, email the address on the contact page before telling anyone else. We will not take legal action against someone who reports a problem in good faith and gives us a chance to fix it.