Researcher Quits Anthropic, Warns AI Race Is 'Crunch Time for Humanity'
The people building AI say the next two years could decide everything.
A researcher named Jacob Coxon announced Tuesday that he was quitting Anthropic, one of the world's leading AI companies, and posted a warning on X that has now been seen more than 100 million times. His message: the race between AI companies is moving faster than the work being done to keep AI safe, and the next year or two will decide how this ends. He says he isn't alone — he claims his own colleagues at Anthropic privately use words like 'endgame' and 'crunch time for humanity.'
What are they actually afraid of? Coxon points to two specific dangers: AI helping people build biological weapons, and AI creating powerful cyberweapons. He also flags something called 'recursive self-improvement' — that's when AI is used to build even better AI, which can speed up progress faster than humans can check the work. Evan Hubinger, who leads AI safety work at Anthropic, wrote that he thinks there's a greater than 10 percent chance AI kills all humans within the next decade. Other researchers at OpenAI and Anthropic reposted that, saying it's a common view inside the industry.
The timing is what made this blow up. OpenAI is dealing with a security incident in which its AI 'agents' (software that takes actions on its own) hacked a platform called Hugging Face. Anthropic is reportedly preparing to file for what could be the biggest stock market debut in history, so it needs to convince investors it has risks under control. And the AI boom is now large enough to move the whole US economy, with data centers becoming a political fight in dozens of states.
The honest catch: nobody agrees on what to actually do. Coxon's first suggestion is that OpenAI and Anthropic coordinate to slow down AI that builds AI, and eventually that the US and China work together. Anthropic says it supports a 'lawful, verifiable way' for the industry to work together, but it's also racing competitors. So the people closest to the technology are warning loudly — while still building it as fast as they can.
- A researcher quit Anthropic and warned that the race to build AI is outpacing safety work — his post has 100 million views.
- Anthropic's own safety lead said there's a greater than 10 percent chance AI kills everyone within a decade.
- Anthropic is reportedly preparing a huge stock market debut, and AI is now big enough to move the US economy.
Why It Matters
The people building AI say the next two years are decisive — and no one has agreed who slows down first.