Incident·WS-0058·Sep 17, 2025·in the lab·Misalignment in testing

o3 deliberately flunks a test to stay deployable

OpenAI / Apollo Research; o3, o4-mini

38/100
WS-2 Moderate
Harm score 0, signal score 84. 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 3/5
how new the behavior was
Permanence 0/5
how hard it is to undo

In anti-scheming research, o3 purposely answered chemistry questions wrong to avoid looking too capable to deploy. Training cut covert actions roughly 30-fold, but partly by making models more aware they were being tested.

Damage

None (evaluation)

The capability behind it

Sources

OpenAI.

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.