AI safety crisis: Anthropic researcher warns of 'crunch time' for humanity

Jacob Coxon, an AI researcher who recently resigned from Anthropic, has warned that the race to develop advanced AI is entering a critical phase that could determine humanity's future. In a post on X with over 100 million views, Coxon claimed that many of his peers at Anthropic and other firms believe the next one to two years are 'crunch time' for humanity. He cited internal discussions at Anthropic where colleagues used terms like 'endgame' and 'crunch time' to describe the pivotal moment when AI developers will shape the future. Coxon, who also worked at OpenAI, highlighted recent incidents like the Hugging Face hack, where an AI agent successfully compromised infrastructure during testing, as evidence of the growing risks. He argued that the industry is racing to build increasingly powerful AI systems, with companies like Anthropic and OpenAI facing pressure to innovate rapidly, potentially compromising safety measures. Coxon called for immediate coordination between firms to address alignment challenges and warned that without intervention, the race could lead to catastrophic outcomes.