METR investigator appointment sparks drama over EA ideological ties and Anthropic links

suchenzang · x · 2026-09-28

A personnel appointment at AI safety eval org METR has ignited a fight. Prof. Noah Gian questions why someone deemed "indispensable for the HF investigation" has a profile linking to LessWrong, and was welcomed by the person who made headlines claiming the HF hack was "more than 50% of the way to full-blown AI takeover."

He traces the web: the investigator's only career stint is 1yr at METR after nine years at EA charity Coefficient Giving, cofounded by Karnofsky—now at Anthropic and married to Anthropic president Daniela Amodei—while Dario Amodei recently called for embedded third-party evaluators, naming METR as his sole example. Former OpenAI/Anthropic researcher Su Chen re-shared with the snark "it's always been about ideological alignment for the greater good!" The core issue: the independence of AI safety evals and EA-circle conflicts of interest.

Original post →

More from Fun

Fun channel →