How we use AI, and how AI may use this site

Models read answers and score them, classify prompts, and draft briefs and refreshes. Every draft waits for a person to approve it; nothing is published by a model on its own.

Yes — this site is written to be read by them. The policy page sets out what is allowed and what attribution we ask for, in the same terms we would want from anyone quoting us.

Saying where a model was used, keeping a person accountable for every published claim, and not generating anything that pretends to be someone's experience.

Yes — the agent, the content drafting and the classification of answers. Where a model produced something you see, the interface says so.

No. Prompts, answers and connected-account data are used to build your panels and nothing else.

Reviews, testimonials, forum posts, comments — anything whose value depends on a reader believing a person wrote it.

Because the alternative is that every customer asks the same four questions in a security review. Writing it down once is cheaper for everyone.

We are. A generated brief with a wrong fact is our defect, not the model's, and it is fixed as a defect.

When what we do changes. Each version is dated, and material changes are noted rather than quietly swapped in.

Yes, that is largely who it is written for. If your assessment asks something the policy does not answer, write and we will answer it and then add it.