A young researcher who spent three years at the frontier of artificial intelligence has walked away from Anthropic, offering a rare and unguarded account of what he believes is known inside these organizations but rarely spoken aloud in public. Jacob Coxon, 27, is not the first to leave such a company citing safety concerns, but his departure carries a particular weight: he claims the very people building these systems privately fear they may be building something that ends human civilization, even as they continue to build it. His resignation is less a career decision than a moral statement —
Anthropic Researcher Resigns Over AI Existential Risk Concerns
The people building AI earnestly believe it could kill us all by the end of the decade
Mark
So this researcher quit both OpenAI and Anthropic. What made him decide to speak up now, and so publicly?
Mimi
He said he couldn't stay silent anymore. Three years of watching the same pattern at two different companies—the gap between what leadership says in public and what they actually believe in private. That gap became intolerable to him.
Luke
But we should be careful here. We have his account of what senior people believe privately. We don't have those senior people confirming it. He's claiming firsthand knowledge, but we're taking his word for the content of private conversations.
Mimi
That's fair. But he's also putting his name and reputation on it. He's not anonymous. He worked there. He has credibility to lose.
Mark
What does he think will happen if things don't change?
Mimi
He's saying AI labs are building systems that could kill humanity by the end of this decade. That's his core claim. Not maybe, not theoretically—he's saying the people building these systems believe this is possible.
Luke
Again, that's his characterization of what others believe. We don't have quotes from those executives or researchers saying they think extinction is likely by 2035. We have Coxon saying they think it privately.
Mark
Does it matter that he quit twice? OpenAI and then Anthropic?
Mimi
It suggests the problem isn't unique to one company. He's saying it's systemic—the race is happening everywhere, and safety concerns aren't winning.
Luke
Or it could mean he has a particular view about AI risk that he carried with him, and when he didn't see it reflected in company priorities, he left. That's also a valid reading. We don't know if his concerns are widely shared inside these organizations or if he's in a minority.
Mark
What happens next?
Mimi
That's the question. He's made a public accusation. The companies will probably respond. Whether it changes how they operate—that's unknown.
Luke
And whether his claims about what executives privately believe can be verified or tested—that's also unknown. This is a resignation statement, not an investigation.
The Pulse
- A researcher with firsthand access to the inner workings of two leading AI labs is publicly alleging that the race toward superintelligence is being run by people who privately fear it could end human life within a decade.
- The most unsettling charge is not incompetence but awareness — that senior figures at these companies understand the existential stakes and continue accelerating anyway, while carefully moderating their language for public audiences.
- Coxon's second resignation in a row from a frontier AI lab suggests the discomfort he describes is not specific to one organization's culture, but something embedded in the structure of the industry itself.
- Neither Anthropic nor OpenAI had responded to his claims at the time of his statement, leaving his account uncontested in the public record and the internal tensions he describes unaddressed.
- The central question his departure raises is whether individual acts of conscience from people with direct knowledge can meaningfully alter the momentum of an industry moving at this speed.
A young researcher who spent three years at the frontier of artificial intelligence has walked away from Anthropic, offering a rare and unguarded account of what he believes is known inside these organizations but rarely spoken aloud in public. Jacob Coxon, 27, is not the first to leave such a company citing safety concerns, but his departure carries a particular weight: he claims the very people building these systems privately fear they may be building something that ends human civilization, even as they continue to build it. His resignation is less a career decision than a moral statement — a refusal to remain silent about a gap between private knowledge and public language that he found impossible to sustain.
On a Wednesday morning, Jacob Coxon posted a resignation statement to social media and named the reason without softening it: the companies building the world's most powerful AI systems are moving recklessly fast, and the people inside those organizations know it. Coxon, 27, had spent three years working on pretraining systems at Anthropic — and had left OpenAI before that — giving his words the texture of someone who had watched this pattern repeat across institutions, not just within one.
His central claim was not that AI development was dangerous in the abstract, but that the researchers and executives leading it privately believe it could end human civilization within a decade. He said he had direct knowledge of senior figures who held this fear, and that the same people who expressed it behind closed doors chose far more measured language when speaking to journalists or the public. "I hear the same people express fear privately," he wrote, framing the gap between private belief and public statement as the core of what troubled him.
Coxon placed AI development in a category of risk he considered unlike anything else humanity currently pursues — not equivalent to war, pandemic, or climate change, but occupying a different tier of consequence entirely. He offered no policy prescription and proposed no solution. He simply stated what he believed to be true and named the companies responsible.
What remained unresolved was whether his departure would register as anything more than a data point in a debate that has been running for years. The companies had not responded publicly. The momentum toward more capable systems showed no sign of slowing. His resignation raised the question that has shadowed AI safety discourse for some time: whether internal warnings from people with genuine knowledge of these systems are capable of changing anything, or whether the industry's trajectory has already acquired a gravity of its own.
Jacob Coxon, a 27-year-old researcher who spent three years working on pretraining systems at two of the world's most prominent AI companies, walked away from Anthropic on a Wednesday morning with a stark warning posted to social media. He had quit OpenAI before. Now he was leaving Anthropic too, and this time he was naming the reason plainly: the companies building the most powerful artificial intelligence systems on Earth are moving too fast, taking risks they should not take, and the people inside those organizations know it.
In his resignation statement, Coxon said the labs are "racing straight to self-improving superintelligence and gambling with our lives." The phrasing was direct, unadorned by the hedging language that typically surrounds discussions of AI risk in public forums. He went further, claiming that the researchers and executives developing these systems believe artificial intelligence could end human civilization within a decade. This was not speculation on his part—he said he had direct knowledge of senior figures at these companies who held this fear.
What struck Coxon most was the gap between what he heard in private conversations and what the same people said when speaking to journalists or the public. Executives and senior researchers, he suggested, carefully calibrate their language for external audiences, choosing words that sound measured and responsible. Behind closed doors, the tone shifts. The worry becomes visible. "This is not a marketing stunt," Coxon wrote, as if anticipating the dismissal his words might receive. "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."
The resignation placed Coxon in a small but growing category of people who have worked at the frontier of AI development and then stepped away, citing safety concerns. His departure from Anthropic followed an earlier exit from OpenAI, suggesting a pattern of discomfort that persisted even as he moved between organizations. Both companies have positioned themselves as leaders in AI safety research, yet Coxon's claim was that neither was acting responsibly in the way that mattered most—by slowing down.
He made one more assertion that carried weight precisely because of its simplicity: "No other human activity poses this level of danger." The statement did not argue for a particular policy or regulatory framework. It did not propose a solution. It simply placed AI development in a category of risk unlike anything else humanity currently pursues. Wars, pandemics, climate change—these were grave threats, but Coxon was suggesting that the race to build superintelligent systems occupied a different tier of consequence.
The timing of the resignation, announced via social media rather than through formal channels, suggested Coxon had decided that the usual protocols of corporate departure were insufficient for what he felt compelled to communicate. He was not negotiating a severance package or protecting his professional reputation through careful silence. He was speaking, and he was naming the companies by name.
What remained unclear was whether his departure would change anything inside Anthropic or OpenAI, or whether it would be absorbed as one more data point in an ongoing debate about how fast AI should advance and who should decide. The companies themselves had not responded publicly to his claims at the time of his statement. The question hanging over the moment was whether internal warnings from people with direct knowledge of these systems would prove sufficient to alter the trajectory of development, or whether the momentum toward more capable AI would continue regardless.
Notable Quotes
They are racing straight to self-improving superintelligence and gambling with our lives.— Jacob Coxon, in his resignation statement
Many executives and senior researchers will couch their phrasing in the press to sound sensible—but I hear the same people express fear privately.— Jacob Coxon
Want the full story?
Read the original at hindustantimes.com ↗
Technology