Jacob Coxon, a 27-year-old former AI pretraining researcher at Anthropic and OpenAI, has resigned, claiming that the companies are not acting responsibly in developing AI systems. His departure was first reported by The Wall Street Journal, which described it as one of the first examples of an Anthropic employee leaving over AI safety concerns.
In a post on X, Coxon said future AI systems could become capable of hacking, accelerating research, and acquiring resources at speeds humans would struggle to match. He also said people working on advanced AI privately take the possibility of catastrophic outcomes seriously.
He further stated on X, “Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources. We have all witnessed the progress in each of these domains, and progress is not slowing.”
Coxon argued that the central issue is not AI intelligence itself but the incentives driving development. There is pressure on Anthropic, OpenAI, Google, Meta, and other businesses to keep developing AI systems. Additionally, governments have a strategic stake in maintaining the competitiveness of domestic businesses. Voluntary slowdowns may be challenging in that setting.
He said OpenAI employees may not fully recognize the scale of the risks. At the same time, Anthropic understands those risks but feels compelled to remain in the race, fearing that less responsible competitors could move ahead first.
Coxon warned, “A common response is “if they truly believe this, why are they still building it?” At OpenAI, many have not deeply internalized the civilizational stakes. At Anthropic, the stakes are well-understood, but they are locked in a race to get there first – they believe no one else will act responsibly, so they must do it themselves, despite the risk.”
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. More thoughts below.
— Jacob Coxon (@hilbertspaess) September 9, 2026
Coxon argued that AI differs from ordinary business competition because the risks may extend beyond individual companies. A failed effort to build a better search engine may be costly, but its consequences are usually limited. More advanced autonomous AI systems could present broader challenges if developers do not fully understand their behavior.
He also questioned whether private companies should be responsible for decisions with potentially far-reaching consequences. He argued that society may need stronger oversight for technologies that could have transformative effects.
Coxon argued, “Accepting this race and entering the “endgame” is a hubristic gamble that should not be launched from a private company’s Slack. Attempting to speedrun alignment should require extraordinary confidence that there are no better trajectories available.”
Recently, more than 1,100 AI insiders signed an open letter called Pacing the Frontier, which called on the US government to support global initiatives aimed at slowing the advancement of sophisticated AI. In order to handle increasingly automated AI research, the petition asks for the development of technical protections and governance mechanisms. Signatories contend that as AI capabilities develop, careful pacing is required to provide time for safety precautions, monitoring, and risk mitigation.
Coxon said many of the people developing advanced AI genuinely believe it could pose an existential risk by the end of the decade. According to him, AI executives and senior researchers often use more measured language in public, but express deeper concerns in private.
Coxon warned, “The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible – but I hear the same people express fear privately. No other human activity poses this level of danger.”
Despite those concerns, Coxon said he remains optimistic about the possibility of coordination across the industry. He argued that incidents such as the Hugging Face attack have made it more politically feasible for US AI labs to agree to slow development. However, he said the world is not currently on a path to avoid a broader global AI race, which could require significant measures, including a temporary ban on improving model capabilities.
Coxon also directed a message to researchers working in AI labs. He urged them to consider the implications of the next stage of development. He questioned whether increasingly capable reinforcement learning systems should be launched without a rigorous understanding of how they function.
Coxon added, “If you are a lab researcher, I urge you to consider what the next few years will actually feel like. Do you want to kick off a superintelligent RL run without a rigorous understanding of its mind? Should you put your head down because “it’s happening anyway” – or take this moment to call for different conditions?”
The timing of Coxon’s resignation drew additional attention because it came days after OpenAI Chief Scientist Jakub Pachocki warned in an essay titled “An Alien Mind.”
In the essay, the OpenAI chief scientist argued that AI development is entering an area where existing safety techniques may not be enough. He wrote that researchers have already seen signs of systems that could become significantly more capable than humans and argued that society is not prepared for the effects of continued rapid advances in AI. He did not call for abandoning AI development, but said the current pace requires greater caution.
In the essay, he argued that alignment and monitoring remain unresolved and that indefinite scaling at maximum speed is not justified. He also suggested voluntary slowdowns should become more common until shared safety standards are in place.
