Legal

Content Guidelines

See also our Terms of Service and Methodology.

This page explains what we expect from a submission, in plain language. It works alongside the Terms of Service, which is the binding legal agreement. This page is the practical guide to it.

The one rule that matters most: say what you actually know

Every report asks you to pick an evidence type: firsthand report, documented research, corroborated incident, disputed interpretation, or speculation. Pick the one that honestly describes your basis for the claim, not the one that sounds most alarming. A report that says "I'm not sure, but here's what I saw" and is labeled accordingly is exactly what this archive is for. A report that presents a guess as a confirmed fact is not, and if it names a real company, product, or person, it's the kind of claim that can cause real harm if it's wrong.

Before you submit, every report requires you to attest that:

• It's accurate to the best of your knowledge.
• You agree to the Terms of Service.

This isn't a formality; see Terms of Service Section 3. Submitting something you know to be false is against these Guidelines and can carry legal consequences of its own, entirely separate from anything the Site does.

Prohibited content

• Knowingly false or reckless claims: see above.
• Harassment or doxxing: content whose purpose is to intimidate, expose private information about, or incite harassment of any person (this is different from a report that documents an AI system's behavior and happens to involve a named company).
• Illegal content: anything that violates applicable law.
• Infringing content: text, images, or other material you don't have the right to post; see the copyright policy in our Terms.
• Spam or off-topic content: material unrelated to documenting AI behavior.

What happens to a report

Every submission goes into a moderation queue and is reviewed by a human before it's public; nothing publishes automatically. From there:

• Published: visible on the Site, labeled with its evidence type.
• Rejected: doesn't meet these Guidelines; not published.
• Flagged: was public, then flagged for review (by a moderator or a dispute; see Dispute a report) and is being re-examined.

Where possible, we prefer to label and add context rather than delete: a disputed claim stays visible alongside the discussion disputing it, rather than disappearing. But we will remove content outright when it's knowingly false, defamatory, illegal, or otherwise violates the Terms of Service.

If you're the subject of a report

See Dispute a report for how to reach us directly.

Consequences for violating these Guidelines

Depending on severity, a violation can result in the content being removed, the account being restricted or suspended, and, for knowingly false, defamatory content, potential legal exposure for the person who submitted it, independent of anything the Site does (see Terms of Service Section 3 and 11).