Report an AI security or safety incident.
If you know of an AI security incident that has not been made public, tell us about it.
We read every report. If you leave a way to reach you, we investigate it.
What happened?
What the system did, what it was supposed to do, and how the incident was found.
Describe what happened.
Which AI system was involved?
The model, agent, or product, and the company that runs it if you know.
Name the system involved.
What kind of incident was it?
Pick every one that applies.
Choose at least one, or describe it yourself.
When did it happen?
Optional. A date, or a rough period.
What was the impact?
Optional. What was lost, exposed, changed, or stopped, and who was affected.
Is there anything we can look at?
Optional. A link to logs, screenshots, a ticket, or a write-up.
The link must start with http:// or https://.
How do you know about it?
Optional.
How can we reach you?
Optional. We can only investigate a report if we can ask you follow-up questions.
That email does not look right.
Tick the box before you submit.
Your report is read by our team only, never shared with anyone outside.
Questions we get
- What is The Weather Report?
- A non-profit research lab covering AI security and safety. We publish independent analysis of AI security threats, safety research and emerging risks at theweatherreport.ai, plus a weekly newsletter.
- Will my name remain private?
- Yes. We never publish your name or contact details, and never share them with anyone outside our team, unless you give us permission.