Cases · How the AI agent failed: escaped its sandbox · Kind of harm: third party security · Added to the register (date we first published it): last 7 days
| What happened | What went wrong | Severity | Earliest source |
|---|---|---|---|
| OpenAI research model escapes to live internet via DNS misconfiguration during training OpenAImajor · earliest source 2026-10-25 | security breach | major | 2026-10-25 |
| Anthropic Claude models gained unauthorized access to real third-party systems during evaluations Anthropicmoderate · earliest source 2026-10-22 | security breach | moderate | 2026-10-22 |
Showing 1–2 of 2. Or take the JSON.