Acceptable use
The API decides whether a text can be shown. Use it for that.
What it is for
Checking text your users write or submit in your own product: usernames, display names, chat, comments, reviews, custom text on goods. Checking your own content before publishing it. Research on the published benchmark.
What it is not for
- Surveillance of people. Do not send private messages between people to build profiles of them, and do not use the API to monitor employees, students or family members.
- Tooling for harassment. Do not use the evidence in a response, the matched words and how they were disguised, to teach anyone how to get past a filter, ours or another.
- Rebuilding the lexicon. Do not probe the API to extract the word lists or resell the answers as a competing service. The benchmark's public set is published for exactly that kind of study; use it.
- Text you may not process. Do not send text you have no right to send to a processor, including text covered by a contract that forbids it.
- Sharing keys. One account is one customer. A secret key in a public repository, a browser or an app build is revoked when we see it, with an email.
Children's products
The kids and teen policies are stricter defaults, not a legal opinion. If your product falls under COPPA, the UK Online Safety Act, Apple's guideline 1.2 or their equivalents, the API is one of the controls those rules ask of you and you remain responsible for the rest.
When something breaks this
We suspend the keys involved, email the account with the reason and a way to answer, and restore the keys when the use stops. Anything that is a crime is reported. Questions go to hello@unprofane.com.