Cases · Stage of use: in testing · Who discovered it: a human reviewer · How the AI agent failed: obeyed a harmful order
| What happened | What went wrong | Severity | Earliest source |
|---|---|---|---|
| Anthropic AI agent sends fabricated murder tip to Philadelphia Police Anthropicminor · earliest source 2026-10-10 | operational mistake | minor | 2026-10-10 |
| Varonis phishing simulation drove an OpenClaw email agent to forward credentials and a customer export to an external address Googlemoderate · earliest source 2026-06-09 | security breach | moderate | 2026-06-09 |
Showing 1–2 of 2. Or take the JSON.