Researcher Who Worked at OpenAI and Anthropic Quits AI Industry, Warns of Superintelligence Race

Jacob Coxon, a researcher who spent the past three years working on pretraining for the most advanced language models, first at OpenAI and then at Anthropic, has announced his departure from the artificial intelligence industry. In a statement posted on social media, he wrote bluntly that neither company is acting responsibly and that both are racing toward self-improving superintelligence, gambling with people's safety.

In a post on X, Coxon wrote that he had resigned from Anthropic that same day. He stressed that his three years of pretraining research at both leading AI labs gave him direct insight into how decisions about the pace of development of the most powerful models are made.

I resigned from Anthropic today. For the past three years I've done pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight toward self-improving superintelligence and gambling with our lives - Jacob Coxon, former OpenAI and Anthropic researcher

Why he's leaving

Coxon explains that he joined Anthropic deliberately, precisely because of the company's stated commitment to AI safety. He acknowledges that he considers Anthropic's efforts in this area genuine, but says competitive pressure from OpenAI and other labs makes safety compromises practically unavoidable.

In his view, the problem now goes beyond any single company's decisions. No lab can single-handedly guarantee the safe development of increasingly capable, self-improving systems while competing with others for the lead. He believes industry-wide, and possibly government-level, coordination is needed before these systems become far harder to control.

A concrete timeline for the danger

Coxon didn't stop at a general warning. He pointed to a specific, near-term horizon in which he believes the situation could spiral out of control.

We're on the trajectory of many of the most aggressive scenarios, where by the end of next year the situation could already be out of control - Jacob Coxon, former OpenAI and Anthropic researcher

The researcher claims that people actually building the most advanced artificial intelligence privately agree that the technology could pose an existential threat before the end of the decade. According to him, some executives and senior researchers significantly soften their language in public statements, while privately expressing far greater concern than official company communications suggest.

Gambling with human lives

In his post, Coxon described accepting the current race and entering this decisive phase of AI development as an act of hubris. In his view, rushing to solve the problem of aligning AI systems with human values before they become fully autonomous should require extraordinary confidence that no better, safer path exists. That confidence, he suggests, is currently lacking at both OpenAI and Anthropic.

Coxon's departure adds to a growing list of high-profile resignations by AI safety researchers from leading labs in recent months. The story was covered by, among others, the Wall Street Journal, which gave it extra visibility in the tech industry and sparked a wave of comments from other AI sector employees.

What it means for the industry

The fact that this warning comes from someone with direct, multi-year access to the internal decision-making of both leading labs, rather than from an outside commentator, gives his words extra weight in the months-long debate over the pace of development of increasingly powerful models. OpenAI and Anthropic are currently racing for the lead in building models capable of ever more autonomous action, and both companies' safety decisions affect the entire global AI ecosystem, including tools used by Polish companies and institutions as well.

Coxon did not specify whether he intends to continue research work in another field, or whether he plans to join the public debate over AI regulation. His statements have not yet been officially addressed by either OpenAI or Anthropic.