Article · Geopolitics · Technology
An Anthropic researcher quit warning about the race. What actually changed in five days?
The words changed, the rules did not. Anthropic's chief executive proposed slowing capability gains and letting outside evaluators in, OpenAI and Musk agreed, but no commitment carries a date, and the researcher who quit now says the companies should regulate themselves for the time being.
On 9 September Jacob Coxon, who spent three years doing pretraining research at OpenAI and Anthropic, announced he was leaving Anthropic. Within five days his post had reached the US Congress, and the heads of the same companies were saying in public something they had not said before: slow down.
This piece separates three things that headlines merge: what Coxon claims, what happened next, and what of it actually binds anyone.