AI Development

Former Anthropic safety researcher Joe Benton warns of AI race risks

September 12, 2026 · Medium importance · policy

Joe Benton stated he left Anthropic’s safety team and would join METR for independent evaluations, warning that frontier labs are racing toward systems smarter than humans while underinvesting in safety. Indian reporting linked the departure to the broader wave of researcher exits and public risk warnings. The move adds external evaluation capacity and public pressure for transparency on alignment progress.

Date
September 12, 2026
Importance
Medium 3/5
Category
policy
Axis of Change
Regulatory Constraint
Organizations
Anthropic METR
Models Affected
No specific model identified
  1. 01 business-standard.com