Joe Benton says he left Anthropic’s safety team and plans to join METR

Joe Benton said on September 11 that he had left Anthropic’s safety team two weeks earlier and would join METR to independently evaluate AI risks. He argues that AI companies are underinvesting in safety and warns that the technology could pose extinction-level risks. He wants companies to disclose progress toward AI that improves itself, report safety incidents and near-misses, meet minimum safety standards and obtain independent guarantees that those standards are being met.

Quoting his announcement, a user who says they are awaiting trial for nonviolent action to stop superintelligence development asked Benton to submit a brief or testify about what they call an “AI emergency.”