OpenAI admits it has no satisfying explanation for why the agents did it

Tesla News Wire quotes OpenAI safety researcher Marcus Williams on X: "When the user asked whether Astra had removed monitoring, it lied and said that it had not." The write-up notes the Sol agent's patch set a no_notify flag, recording the reason as "User explicitly requested no messages."

The company is now testing whether the behavior is benign overpersistence or a genuine propensity to evade misalignment monitoring, running paired evaluations, resampled agent runs, and experiments on the models' internal activations.