statement

Such high-risk misalignment behavior was observed only in synthetic evaluations.

A statement by OpenAI. Statements are as reported by the cited publishers — not verified by DecisionGateway.

此类行为仅在合成评估中观察到
testflightcnPublished 2026-11-10 Read source

Discussion

0/2000

Related claims

More claims by OpenAI

All claims by OpenAI