A colleague used AI output and something genuinely went wrong, now what?

AI incidents look like other incidents: contain fast, limit damage, learn. The difference is the cause often sits in a prompt or hallucination, not in a log with a stack trace. Treat it as a production incident, not a personal mistake.

Try this first

1. 1Stop work in flight: the email not yet sent, the quote not yet out, the code not yet in production. First stop spread, then analyse.
2. 2Collect evidence: the prompt, the output, the tool, the model version, the timestamp. Screenshot or copy the chat log before it gets overwritten.
3. 3Assess impact: did customer data leak, was a wrong invoice sent, was a wrong diagnosis or advice given? On a data breach the 72-hour AP notification clock starts.
4. 4Communicate where needed: the customer or vendor gets a short correction, internally the team gets a no-blame summary of what happened.
5. 5Capture the lesson in the AI policy or tools matrix. A repeat of the same pattern is a process failure, not a person failure.

None of the above fits?

Describe your situation below. We pass your input plus the steps you already saw to our AI and return tailored next-step advice. If it's too risky to DIY, we'll say so.