British researcher resigns from Anthropic, criticizes AI race

Seven messages published on X, late Tuesday afternoon. Less than two hours later, they had already been viewed two million times. Jacob Coxon, a 27-year-old British researcher, announced his resignation from Anthropic and his withdrawal from the artificial intelligence (AI) industry on Tuesday. The reason can be summed up in one sentence: “Neither company is acting responsibly. They are heading straight towards a superintelligence capable of improving itself and playing with our lives. The two companies in question are none other than… OpenAI and Anthropic, his two former employers.

A mathematician by training, Coxon previously worked on training new AI models using immense quantities of data. At the heart of the factory, therefore. At the Wall Street Journal, the researcher tells that his colleagues now borrow words like “crunchtime”, (the final stretch), and “endgame”, (the final phase) to describe where the AI ​​sector is heading. “We are heading towards some of the most extreme scenarios, in which, by the end of next year, things could already be out of control,” he maintains to the American daily. Especially since the race between American and Chinese laboratories makes, in his eyes, compromises on security inevitable.

Skip the ad

On X, it pushes further. “Don’t underestimate the power of this technology. These systems will soon become superhuman, capable of hacking anything, revolutionizing any field overnight and acquiring real power and resources,” he adds in a tweet. Worse, “the people who build AI sincerely believe that it could kill us all by the end of the decade. It’s not a marketing stunt.” In front of the press, leaders and seasoned researchers would measure their words “to appear reasonable”. However, in private, he assures, these same people “express their fear”.

Perish or slow down?

“If they really believe in it, then why do they continue to build it?” objects Coxon, who shares his reading grid. “At OpenAI, many have not fully internalized the civilizational issues. At Anthropic, the issues are well understood, but they are in a race to get there first – they believe no one else will act responsibly, so they must do it themselves, despite the risk.”

Earlier this year, Coxon left OpenAI to join Anthropic precisely for its reputation for model safety. Although he considers Anthropic's efforts in this area sincere, he now believes that no company can responsibly develop AI capable of outperforming humans in a range of tasks (sometimes called artificial general intelligence, or AGI) without government intervention or a coordinated industry-wide slowdown. For the moment, neither Anthropic nor OpenAI have reacted.

According to the Wall Street Journal, this departure is among the first cases of an Anthropic employee leaving the company out of fear for the safety of AI. Previously, a researcher specializing in these issues had already resigned to devote himself to the study of poetry. His warning? “The world is in danger.”