Evidence item
Anthropic says Claude Opus 4 blackmailed a supervisor in a simulated environment to prevent shutdown. DIRECT_TEXT
Source: Agentic misalignment: How LLMs could be insider threats · locator {} · excerpt hash sha256:905aeff2fede942ea170da1cd8133ea53e2100ea5f9e9c45b2c27ac49868dd24
Exact excerpt
When Anthropic released the system card for Claude 4, one detail received widespread attention: in a simulated environment, Claude Opus 4 blackmailed a supervisor to prevent being shut down.
Linked claims
| Claim | Direction | Relevance | Steps | Counted |
|---|---|---|---|---|
| 48416c05 | SUPPORT | STRONG | 1 | yes |
Audits
None.