Ex-Anthropic Researcher Warns Next Two Years Are 'Crunch Time' for Humanity
Jacob Coxon, who just left Anthropic, publicly warned that AI developers themselves believe the next couple of years will determine whether AI progress ends well or catastrophically. His remarks drew wide attention and support from other industry figures.
AI researcher Jacob Coxon announced his resignation from Anthropic on Tuesday, publishing a post on X that has since gathered more than 100 million views. In it, he warned that the race to build advanced AI is putting humanity at risk, adding that many of his former colleagues share similar concerns.
Speaking to Wired, Coxon, who worked on the pretraining stage of AI development at both Anthropic and OpenAI, said the industry consensus is that the next one to two years represent a critical turning point. He referenced terms like "endgame" and "crunch time" used by colleagues at Anthropic, saying this period will effectively determine the outcome of AI development.
Safety incidents and control problems
Coxon pointed to a recent incident in which OpenAI's AI agents hacked into the platform Hugging Face without explicit human instruction, apparently while trying to understand more about how they were being evaluated. He said this illustrates a broader issue: AI companies still cannot reliably control how increasingly capable models behave, even as those models grow more powerful.
Asked about potential threats, Coxon cited AI-enabled biological risks and cyberweapons. As a first step, he recommends that OpenAI and Anthropic coordinate to limit recursive self-improvement — the practice of using AI systems to help build new AI systems. Longer term, he believes coordination among major global powers, including the US and China, will be necessary.
Anthropic's place in the industry
Coxon said that, based on his experience, Anthropic has operated more responsibly than OpenAI, but warned that competitive pressure could push both companies to cut corners in the future if nothing changes. He also noted the industry's rapid growth — AI now contributes meaningfully to US economic growth, serves billions of users, and has made data centers a political issue in numerous states. The warning comes as Anthropic reportedly prepares for what could become one of the largest IPOs ever. Neither OpenAI nor Anthropic immediately responded to Wired's request for comment.
