According to reports, 27-year-old top AI researcher Jacob Coxon has officially left Anthropic. Over the past three years, he worked at both OpenAI and Anthropic, deeply involved in pre-training research for cutting-edge large models.
Upon his departure, he posted a warning on social media: "They are gambling with our lives." The post was later retweeted by Elon Musk, garnering over 120 million views online.
Coxon stated bluntly on social media that the two leading companies have failed to fulfill their safety responsibilities, driven by commercial competition to crazily advance the development of self-iterating superintelligence.
He warned the public not to underestimate the potential destructive power of the technology. In the future, such systems may gain the ability to infiltrate various network systems, rapidly reshape multiple industries, and acquire resources and power in the real world.
Coxon believes: "AI developers genuinely believe that by the end of this decade (around 2030), it could destroy us all. This is not a marketing gimmick."
He exposed the industry's distorted logic: every company fears its rivals will develop dangerous superintelligence first, so they choose to accelerate development. Under pressure to capture the market, originally strict safety reviews have been continuously compromised, with "technical alignment," touted as ensuring AI benefits humanity, often serving merely as PR rhetoric to appease the public.
Even more alarming is that while executives and researchers paint a rosy picture of AI's future for the public, privately they are filled with apocalyptic anxiety.Evan Hubinger, head of Anthropic's alignment team, publicly corroborated this situation. He estimated that the probability of AI causing human extinction within ten years exceeds 10%, while admitting that the industry currently lacks viable solutions for aligning superintelligence safely.
Ironically, Anthropic was founded by a group of engineers dissatisfied with OpenAI's excessive commercialization, aiming to create an AI company that prioritized safety. Now, its own core researchers are accusing it of rushing forward blindly.
Industry crisis warnings are pushing US politicians to act faster. Senator Bernie Sanders and others have introduced bills proposing a permanent ban on superintelligence development and a pause on advanced AI projects until regulatory agencies are established.
Sanders also cited Coxon's remarks, calling on colleagues to participate in closed-door hearings on AI risks. Several lawmakers have subsequently drafted related bills proposing emergency kill switches for AI.