For years, warnings about artificial intelligence destroying humanity largely came from philosophers, researchers, science-fiction writers and a handful of technology insiders.

This week, something felt different.

The warning came from inside the labs.

And it didn’t come in the form of another open letter signed by hundreds of people.

It came from a researcher who actually walked away.

The resignation that got everyone’s attention

Jacob Coxon, a researcher who had worked at both OpenAI and Anthropic, resigned from Anthropic this month and publicly explained why.

His message was unusually blunt: he believed the companies developing frontier AI were moving too quickly and were not taking the potential risks seriously enough.

More importantly, Coxon reportedly left just two months before his Anthropic equity was due to vest.

That’s what made the story different.

People can dismiss an academic paper. They can ignore another technology executive warning about the future. They can debate an open letter.

But when someone working inside one of the world’s leading AI laboratories voluntarily walks away from money and a prestigious position because he believes the technology is moving in a dangerous direction, people naturally start paying attention.

Coxon has also put a number on his concern: he has estimated roughly a 10% chance of AI causing human extinction within the next decade. That is his assessment, not an established scientific probability—and there is no consensus among experts on such a number.

Then something even more unusual happened.

One of his colleagues didn’t dismiss him.

Evan Hubinger, who works on alignment research at Anthropic, publicly agreed with the broader concern and said he believes AI could potentially kill humanity.

That made the story much harder to treat as simply one former employee’s opinion.

Then the CEOs started talking about slowing down

And this is where the story becomes genuinely interesting.

Anthropic CEO Dario Amodei publicly argued that the pace of frontier AI development needs to slow down and called for stronger safeguards and independent evaluation.

He raised an extraordinary possibility: that swarms of increasingly capable AI agents could potentially take over significant parts of the internet within six months to a year if their capabilities advance faster than the safeguards designed to control them.

OpenAI CEO Sam Altman subsequently agreed that frontier AI development needs to be paced more carefully and supported the idea of independent evaluators having meaningful access to AI systems.

Think about the contradiction.

The companies competing to build increasingly powerful AI are simultaneously telling us that the technology may be advancing too quickly.

That is not a contradiction that can be easily laughed away.

But it is also not proof that an AI apocalypse is coming.

And that distinction matters.

Nobody actually knows how big the risk is

This is perhaps the most important part of the conversation that often gets lost in the headlines.

There is no scientific consensus on the probability that AI will cause human extinction.

Some researchers believe the risk is signific