Coxon quit. The rest of the building started talking.
A former Anthropic researcher left midweek saying labs are gambling with lives. By Friday, safety staff at Anthropic and OpenAI were saying the quiet parts on the record.
Feature1 minIn-depth
Priya Shah1 min read
- Anthropic
- OpenAI
- Safety

Photo: Business InsiderSource page
Jacob Coxon resigned from Anthropic this week after three years on pre-training research. He had previously worked on GPT-4o at OpenAI. In public remarks picked up by Business Insider, he said leading labs are gambling with our lives. Resignations happen. This one became a permission structure. Staff who had been posting in a personal capacity suddenly had a news peg.
Hubinger’s 10 percent, Anna Wang’s “look up,” Jasmine Wang’s RSI warning, and Marcus Williams on OpenAI’s safety oversight bench all landed in the same 72-hour window. Williams’s line, as Insider summarized it, was that without regulation or a coordinated slowdown the risk stays on the table. None of that is a joint statement. It is a cluster. Clusters are how a beat decides the week had a subject.
The morning pieces on Musk calling the chorus a psyop, and on Altman being open to slowing, are still the right frame for motive and for markets. The evening add is personnel. A named departure plus a stack of named researchers is harder to dismiss as mood. It is also not a policy. Coxon is not on the board. Hubinger does not set the cluster schedule.
What to watch next is whether anyone inside either lab follows him out, and whether the IPO calendars slip by a week that can be blamed on “process.” If the only sequel is more posts, this was a news cycle. If a training run moves, it was a resignation that did work.


