September 9, 2026, (Inside AI) — A 27-year-old researcher has publicly resigned from Anthropic, warning that leading AI labs are racing toward self-improving superintelligence without adequate safeguards.
Jacob Coxon, who spent three years in pretraining research at Anthropic and OpenAI, announced his departure in a post on X. His message was blunt: neither company is acting responsibly.
“Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives,” Coxon wrote.
The resignation lands at a moment of intensifying public debate over AI risk. Coxon’s warning is not the first from an insider, but it is among the sharpest from someone so junior in years yet so close to the technical core.
Read: 1,100+ OpenAI, Google, Meta Staff Urge US to Pace Frontier AI
Private Fears Outpace Public Language
Coxon claimed his concerns stem partly from private conversations with senior researchers and executives. He said many use cautious language in public but express far greater fear behind closed doors.
“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,” he wrote.
He argued that advanced AI could eventually hack computer systems, transform entire fields almost overnight, and accumulate real-world power. He went further, saying sufficiently advanced AI could pose an existential threat within several years.
“The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon claimed.
His framing echoes a long-running tension in AI safety circles. Public statements from lab leaders often emphasize caution. Private remarks, according to multiple insiders over the years, can be far more alarmed.
Read: Anthropic CEO Dario Amodei Calls for Pacing AI Frontier Development
Anthropic Knows the Stakes but Keeps Racing
Coxon drew a clear distinction between his two former employers. He said Anthropic understands the civilizational stakes better than OpenAI, but that awareness has not slowed its pursuit of more capable systems.
“At OpenAI, many have not deeply internalized the civilizational stakes. At Anthropic, the stakes are well-understood,” he wrote.
He accused Anthropic of being trapped in a competitive race. The company, he claimed, believes that if it does not build advanced AI first, another organization will do so without adequate safety measures.
That dynamic, he argued, pushes labs to continue advancing capabilities even while acknowledging catastrophic risk. It is a version of the classic collective action problem applied to frontier AI.
His post quickly went viral, drawing both support and mockery. One user wrote, “After years of contemplation, you quit publicly to make a point then use the attention to offer zero solutions or better path forward.”
Another compared the moment to nuclear weapons: “We are building the bomb again only this time, it thinks.”
Coxon offered no policy roadmap or technical fix. But his resignation adds another data point to a growing pattern of insider warnings from young researchers at top labs.
Whether that pattern shifts corporate behavior remains unclear. For now, the race he described shows no sign of slowing.
