Anthropic Researchers Back Extinction Warnings, Citing >10% Decade Risk as Musk Calls It a Psyop
Updated
Updated · The Guardian · Sep 10
Anthropic Researchers Back Extinction Warnings, Citing >10% Decade Risk as Musk Calls It a Psyop
3 articles · Updated · The Guardian · Sep 10
Summary
Several Anthropic researchers publicly backed ex-colleague Jacob Coxon after his resignation, saying advanced AI could threaten human survival within 10 years and that more staff privately share those fears.
Anna Wang said many employees want development slowed because no viable scientific plan exists to control recursively self-improving AI, while Drake Thomas said work is moving too fast for the needed safety assurance.
Evan Hubinger said he personally sees a greater than 10% chance of AI killing all humans within the decade, and Samuel Marks wrote that senior employees tend to be more alarmed.
Elon Musk called the wave of warnings a “setup” and “psyop,” amplifying an evidence-light theory that the posts were meant to build support for heavy AI regulation; Bill Ackman also highlighted the claim.
Anthropic defended its approach as transparent about both benefits and unprecedented risks, and on the same day said it had disrupted an attempt to use its models to build a biological weapon.