Joe Benton
Role at the time: Former Anthropic safety-team member, writing in a personal Substack post.
Challenged the characterization
Responding to
Is Anthropic acting responsibly while advancing toward recursive self-improvement?
“Frontier AI companies are racing to build AI systems that can recursively self-improve.”
Why we used this label
Benton explicitly challenges the characterization that frontier-AI companies are sufficiently pacing and resourcing safety work. His post is personal commentary from a former employee, not independent verification of internal practices.
Joe Benton — Why I left Anthropic’s safety team to hold AI companies accountable (opens in a new tab)Opening paragraphs
Personal post by a former Anthropic safety-team member describing his assessment of frontier-AI incentives and his move to METR. It is attributed commentary from a former employee, not independent verification of company practices.