Jacob Coxon, a researcher who recently left Anthropic, has issued a stark warning about the trajectory of artificial intelligence, describing the current era as “crunch time for humanity.” In an interview with WIRED, Coxon discussed the internal efforts at Anthropic—described as a “mini Manhattan project” focused on safety—and argued that the industry’s commitment to alignment remains insufficient.
Coxon emphasized that AI laboratories have only a limited number of years to ensure their systems are safe before they become too powerful to control. The departure highlights growing tensions within the tech sector regarding the pace of development versus the urgency of safety protocols.
As discussions around AI risk intensify, Coxon’s exit serves as a prominent example of internal dissent, with researchers increasingly vocal about the potential existential threats posed by unaligned advanced models.
If the window is closing, why did he wait until leaving to speak so openly? Transparency now would help more than resignation.
Did anyone else catch the bit about the mini Manhattan project? That analogy seems apt if we treat safety with that same urgency.
Crunch time for humanity? That’s dramatic. I think the risks are real, but ‘existential’ feels like fear-mongering to get funding.
This is terrifying but honestly not surprising. The race to deploy is outpacing safety research.