An artificial intelligence researcher who recently resigned from Anthropic has ignited a fierce global debate by warning that the rapid pace of model development poses an existential threat to humanity. Jacob Coxon, who spent three years doing pretraining research at both OpenAI and Anthropic, announced his departure in a threaded statement on X that quickly amassed more than 156 million views and over 752,000 likes as of September 9, 2026.
Neither company is acting responsibly,
Coxon wrote in his viral post, They are racing straight to self-improving superintelligence and gambling with our lives.
Inside the Risk of Superhuman AI and Recursive Self-Improvement
The warnings from departing researchers center around artificial superintelligence—a hypothetical benchmark where an AI system’s cognitive capabilities vastly exceed human ability across every domain, including science, mathematics, and complex problem-solving. Coxon warned that these systems will soon become superhuman and capable of hacking anything.
Speaking with the BBC, Coxon elaborated on the potential scale of the crisis, noting that fears of an autonomous swarm of bots acting like a supercomputer could become realistic within six months to a year. I believe that if we don’t slow down at the current rate of progress, there is a strong chance that we could all die in the immediate future,
Coxon told the BBC.
“The people who work at these companies are completely serious when they ask for regulation because they find themselves trapped in a race. And they’re scared of the outcomes of that race.”
Jacob Coxon, ex-Anthropic researcher
Internal Confirmation from Current Anthropic Staff on Alignment Risks
Rather than dismissing the alarming claims, other researchers inside the leading AI laboratories stepped forward to confirm that existential threat models are taken seriously internally. Evan Hubinger, an alignment science lead and team leader in AI alignment stress testing at Anthropic, responded directly to Coxon’s posts on X.
Jacob is correct here – we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade,
Hubinger wrote, adding that while current models present low risks, recursive self-improvement poses a severe danger.
Hubinger also noted that although Anthropic is trying its best, the organization does not yet have a definitive plan to solve alignment for superintelligence and is not clearly on track to achieve one.
Commercial Competition and the Push for Global Regulation
The relentless acceleration of model capabilities is primarily fueled by intense commercial rivalry. Both OpenAI and Anthropic are hurtling toward initial public offerings while competing against a swath of international competitors, including cheaper, high-performing AI models originating from China. Anthropic publicly acknowledged earlier this year that it had loosened some of its safety standards to keep up with market rivals.
Shama Hyder, a professor of AI at Link School of Business, observed that every major player believes slowing down simply hands the competitive advantage to someone less careful, creating a systemic trap that forces even cautious developers to accelerate. In response to these pressures, Dario Amodei, head of Anthropic, published an essay proposing a three-point plan that calls for independent model monitoring, industry-wide regulation, and global regulation.
Coxon welcomed the proposed slowdown but emphasized that any effective pause must be coordinated globally. Sam Altman and Elon Musk have agreed that this is a great idea. But I think there’ll need to be some sort of coordinated slowdown with China if we’re going to avoid a race at an international scale,
Coxon told the BBC.
Political Fallout and Growing Pressures from Washington and State Leaders
The public disclosures have expanded far beyond Silicon Valley boardrooms, drawing swift reactions from elected officials. Illinois Governor JB Pritzker took to X to voice alarm, writing that It’s becoming more clear the threat AI poses to humanity, so I’m calling for immediate action from the industry and Washington,
These departures follow a pattern of high-profile resignations across the artificial intelligence sector. In February, Anthropic safety lead Mrinank Sharma stepped down, warning that the world is in peril. Around the same time, OpenAI researcher Zoë Hitzig resigned over ethical concerns regarding profit prioritization. With prominent technical staff walking away from equity and stability to sound public alarms, policymakers face mounting pressure to determine how external oversight can be established before frontier models outpace human control.