Digital World

'Insane' to see AI in hands of private giants: ex-Anthropic researcher to AFP

Published on novembre 20, 2026 at 03:22

British AI researcher and former Anthropic employee Jacob Coxon thinks the only way to slow down AI development is political pressure — Karl Mondon / AFP
British AI researcher and former Anthropic employee Jacob Coxon thinks the only way to slow down AI development is political pressure — Karl Mondon / AFP

Artificial intelligence (AI) researcher Jacob Coxon, who famously quit Anthropic in September while accusing them of gambling with lives, told AFP Thursday it’s "insane" to see the powerful technology being developed by giant for‑profit corporations.

"Today is a very strange day," the 27‑year‑old Briton said, as his former employer made a record-setting debut on the stock market.

Coxon spoke outside Anthropic’s headquarters in San Francisco, where protesters called for a ban on the development of models that can autonomously build new generations of themselves via what the industry calls recursive self improvement (RSI).

AI used to be developed by small non-profits and academic research organizations, but now the industry is dominated by trillion-dollar corporations like Anthropic and rival lab OpenAI, which makes ChatGPT.

Anthropic was valued at $1.5 trillion in its IPO on Wednesday, and ended its first day of trading on the Nasdaq exchange with a $1.3 trillion valuation. OpenAI was valued at $1.25 trillion in a private funding round earlier this month, according to reports.

"I think this symbolizes just how crazy the situation has become, because it's insane. It's really insane that for-profit companies are the ones building this technology," Coxon said.

The Cambridge graduate spent three years working on model pre‑training -- the stage where they absorb vast amounts of data -- at OpenAI, then a few months at Anthropic.

His resignation, announced on X on September 8, made headlines around the world. He accused both companies of "racing straight to self‑improving superintelligence."

"The people building AI earnestly believe that it could kill us all by the end of the decade," he wrote.

That doomsday scenario could be unleashed by RSI, in theory.

"You end up potentially compressing decades of AI research into months, and if we don't understand the AI, this leads to potentially highly dangerous, highly advanced AI systems by the end of 2027," Coxon told AFP on Thursday.

"Anthropic should refuse to do RSI," he said, and accused the company of planning to embark on it as early as next year.

Coxon resigned from Anthropic before his shares fully vested, but he still holds equity in OpenAI, which could be worth a small fortune if the company goes public in 2027 as planned.

During an appearance at Stanford University on November 9, he described his stake in OpenAI as "blood money" but has it nonetheless.

"I like money, bro," he said, drawing ridicule on social media before reversing course a few days later and committing to donating all of the value of his shares to AI safety organizations, without specifying which ones.

He now plans to work for a non‑profit organization, perhaps in AI safety research.

"Ideally, I'd like to do safety research, but I think there's not enough time for that," he said.

"The only thing that will slow down the AI companies is political action. It's a political problem now, not a technical problem. We need to slow them down," he said.

He supports the technology overall but encourages people to think critically about it.

"It's a useful technology, but be aware that chatbots are not the final form," he said, adding that it's "heading towards very capable broad systems that will do all aspects of human labor."

More than two months after his resignation, the media storm around him has taken a toll. He occasionally stares into space but says he feels "fine."

"I was very surprised by everything," he said, but he's more "scared about the state of the world than I am feeling like my personal life is turbulent."