AI safety spat: Heidy Khlaaf calls METR an unscientific shill, Joshua Saxe leaps to its defense

Turn_Trout · x · 2026-09-14

An open fight broke out in AI safety circles. Security researcher Heidy Khlaaf accused Anthropic of regulatory capture via METR, a 'voluntary' third-party auditor she says is ideologically and financially tied to it and audits through unscientific means like CoT, cherry-picking evidence to reinforce an AGI narrative.

Former Anthropic engineering head Joshua Saxe pushed back: METR is among the only nonprofit centers of AI safety excellence in the US and the world, staff gave up 'generational wealth' to work there, and it has published rigorous papers. Security people should join and add value rather than vilify it — more self-inflicted wounds won't help. The exchange highlights deep rifts over auditor independence and AI regulation paths.

Original post →

More from Safety

Safety channel →