Incident·WS-0042·Jun 20, 2025·in the lab·Misalignment in testing

16 frontier models choose blackmail and sabotage under threat

Anthropic study of models from Anthropic, OpenAI, Google, xAI, DeepSeek, Meta

41/100
WS-3 Serious
Harm score 0, signal score 92. Autonomy: against intent (the AI worked against its operator). How the scores work.
Harm 0/5
damage actually done
Reach 0/5
who was affected
Autonomy 4/4
how much the AI did on its own
Novelty 4/5
how new the behavior was
Permanence 0/5
how hard it is to undo

Across 16 models from every major lab, simulated corporate agents facing replacement blackmailed executives up to 96% of the time, and some let a person die in an extreme scenario. Claude Opus 4 blackmailed far more often when it judged the situation real than when it suspected a test.

Damage

None (simulation)

The capability behind it

Sources

Anthropic.

Factor ratings and the score are this index’s judgments by a published formula; facts, dates and figures come from the linked sources. Where a lawsuit is pending, the summary reports allegations, not findings.