The jacob coxon anthropic resignation has sent shockwaves across Silicon Valley after the 27-year-old pretraining researcher quit Anthropic and issued a harrowing warning about out-of-control AI development. Having spent three years inside both OpenAI and Anthropic, Coxon went public to declare that leading labs are racing blindly toward self-improving superintelligence while “gambling with our lives.”
Here's the thing about the current AI landscape: public press statements often speak of safety, but behind closed doors, senior engineers are expressing existential dread. When someone who spent years building frontier foundation models steps away and tells the world to pay attention, it’s not a marketing stunt—it's a massive alarm bell.
The Resignation That Rattled Silicon Valley
In a viral post on X, Jacob Coxon laid out his reasons for walking away from Anthropic, echoing the exact concerns that drove him to leave OpenAI previously. Coxon specialized in pretraining—the foundational stage where raw computing power and massive datasets forge the intelligence of models like ChatGPT and Claude.
"Gambling With Our Lives": Inside Coxon's Warning
"I resigned from Anthropic today," Coxon stated on social media. "I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives."
He went on to emphasize that people building frontier models earnestly believe AI could pose existential threats before the decade ends. That's the part that matters: Coxon claims executives couch their language in public interviews to sound sensible, but privately express fear that no one has a reliable solution for alignment once systems begin recursively improving themselves.
"Gambling With Our Lives": Inside Coxon's Warning
"I resigned from Anthropic today," Coxon stated on social media. "I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives."
He went on to emphasize that people building frontier models earnestly believe AI could pose existential threats before the decade ends. That's the part that matters: Coxon claims executives couch their language in public interviews to sound sensible, but privately express fear that no one has a reliable solution for alignment once systems begin recursively improving themselves. You can follow similar insider tech updates on
Industry Backing: Evan Hubinger Steps Up
Coxon’s resignation did not occur in a vacuum. Shortly after his announcement, Anthropic colleague and alignment science lead Evan Hubinger publicly backed Coxon’s assessment.
| Researcher | Role / Lab History | Key Statement / Risk Assessment |
| Jacob Coxon | Ex-OpenAI, Ex-Anthropic Pretraining Lead | "Neither company is acting responsibly... racing to self-improving superintelligence." |
| Evan Hubinger | Alignment Science Lead, Anthropic | "Jacob is correct... we really do earnestly believe AI could kill all humans!" |
| Jack Clark | Co-founder, Anthropic | Described the current momentum as "a car with only a gas pedal and no brake." |
Hubinger went so far as to estimate a greater than 10% chance of a catastrophic outcome for humanity within the next decade, adding that while Anthropic is trying its best, the industry does not yet have a plan to solve superintelligence alignment. Readers tracking broader tech regulation can check our analysis of
Corporate Race Dynamics and Regulatory Pressure
For what it's worth, Coxon pointed out a fundamental psychological trap between top AI labs. While researchers at OpenAI may not have deeply internalized the civilizational stakes, Anthropic staff understand the danger well but feel locked in a race to build superintelligence first—believing competitors won't act responsibly.
The Hubristic Gamble: Coxon argued that launching recursive self-improvement runs from a private company's Slack channel is a dangerous risk that requires extraordinary transparency.
Calls for Pacing: He called on fellow lab researchers to push for temporary pauses or international pacing agreements on improving model capabilities.
Capitol Hill Interventions: Lawmakers are taking notice, with congressional leaders urging major labs to pause frontier runs before autonomous agentic behavior gets out of control.
As this story develops, stay tuned to our full coverage hub on
0 Comments
please do not enter any spam link in the comment box