We've reached the point where, after 25y in cybersecurity, I feel morally obligated to say this for the record:
The narrative being pushed around AI safety, sandbox incidents, and the suggestion that METR be treated as an authority is dangerous, deceptive, and morally corrupt.