What to do when AI causes harm

A practical response checklist when AI-related harm shows up: scams, account takeover attempts, nonconsensual imagery, workplace incidents, or dangerous advice. Literacy without panic.

Minutes matter for fraud and safety. Knowing the next step beats doomscrolling.

1. Stabilize: If physical danger → emergency services.
2. Preserve evidence: screenshots, URLs, timestamps (don’t destroy devices needed for investigation).
3. Cut access: change passwords, enable MFA, revoke app tokens, freeze cards if financial.
4. Report: platform abuse forms, school/employer, bank fraud line, police when appropriate.
5. Support people: targets need belief and practical help—not blame.
6. Review systems (builders): turn off risky features, patch prompts/tools, notify users if required.

Follow published incident procedures. Do not freelance “investigations” that violate policy or law.

• No retaliation guides, no hacking back, no illegal countermeasures.

⚠️ Myth: If you clicked once, nothing can be done.

✓ Reality: Speed + reporting still reduces damage often.

⚠️ Myth: Only experts can report.

✓ Reality: Platforms and banks expect ordinary users to report.

• Platform safety centers (verify)
• FTC / consumer protection scam reporting in your country
• Related Course 29 chapters: scams-and-fraud; nonconsensual-imagery; jailbreaks-and-injection
• NIST AI RMF: https://www.nist.gov/itl/ai-risk-management-framework ↗