Jacob Coxon, a prominent artificial intelligence researcher, announced his resignation from Anthropic on Tuesday, September 9, 2026, after three years spent at the forefront of AI development, including time at OpenAI. In a detailed thread posted on X, Coxon leveled sharp criticism against the leading AI laboratories, accusing them of a dangerous and unchecked pursuit of advanced artificial intelligence. His departure and public statements have ignited further debate within the technology community regarding the pace and safety of AI innovation.
"I resigned from Anthropic today. 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." Jacob Coxon, Artificial Intelligence Researcher
Coxon's experience spanned both OpenAI and Anthropic, two of the industry's most influential companies, providing him with an insider's perspective on frontier AI development. His core accusation is that these firms are "racing straight to self-improving superintelligence and gambling with our lives." He argued that speed, rather than caution, currently dictates the direction of AI development, a trajectory he believes poses significant risks to humanity.
In his extensive remarks, Coxon urged the public to take the impending advancements seriously, insisting, "Do not underestimate the power of this technology." He described the systems on the horizon as capable of dwarfing anything currently available, potentially functioning as "superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources." Coxon asserted that visible technical gains across the board indicate that this progress is not slowing down.
Perhaps the most alarming claim from Coxon's thread concerned the private sentiments within the industry. He alleged that a genuine belief in catastrophic risk is prevalent among the very individuals constructing these systems. Coxon stated flatly, "The people building AI earnestly believe that it could kill us all by the end of the decade." He emphasized that this concern is sincere and not a "marketing stunt," suggesting that public statements from industry leaders often undergo sanitization, presenting a stark contrast to private conversations. He noted that executives and senior researchers "will couch their phrasing in the press to sound sensible – but I hear the same people express fear privately," concluding that "no other human activity poses this level of danger."
Coxon directly addressed the logical question: if the danger is so real, why do these companies continue to build? His answer highlighted a difference between the two labs he worked for. At OpenAI, he suggested that "many have not deeply internalized the civilizational stakes." For Anthropic, he presented a different rationale, stating the company understands the risks clearly but presses forward due to a belief that rivals will not exercise similar restraint, essentially being "locked in a race to get there first." Both companies are key players in the industry's sprint toward artificial general intelligence (AGI), a milestone signifying machines that can match human ability across virtually any task. Anthropic, in particular, has cultivated a reputation as a safety-conscious counterweight, a reputation that has faced scrutiny in the past.
Coxon did not mince words in his assessment of the industry's current approach, labeling the race toward advanced AI as "a hubristic gamble that should not be launched from a private company’s Slack." He contended that any rush to implement safety solutions "should require extraordinary confidence that there are no better trajectories available."
Despite his dire warnings, Coxon's message was not entirely devoid of hope. He expressed a degree of optimism regarding the potential for industry-wide coordination, citing "warning shots like the Hugging Face attack" as evidence that cooperation among U.S. labs has become more realistic. However, this optimism came with a significant caveat. He cautioned that current momentum is insufficient to prevent a dangerous free-for-all, stating, "I don’t feel like we’re on track to prevent a global race, which may require costly actions such as a temporary ban on improving model capabilities."
Coxon concluded his thread by directly appealing to researchers still working within AI labs, urging them to reflect on the industry's direction. He posed pointed questions about the risks of unchecked development, asking whether they wished to "kick off a superintelligent RL run without a rigorous understanding of its mind," or to "put your head down because ‘it’s happening anyway’ – or take this moment to call for different conditions." His exit adds to a growing, albeit small, number of departures from major AI labs driven by safety concerns, further fueling the ongoing public and private debate over the speed and prudence of AI development.
