Anthropic Alignment Lead Warns AI Could 'Kill All Humans' As ...

The article reports that Jacob Coxon, described as a researcher at Anthropic, has announced his resignation. Coxon is quoted as saying that AI companies are "racing straight to self-improving superintelligence and gambling with our lives." The article further reports that Anthropic alignment lead Evan Hubinger has responded to Coxon. The exchange between the two is the central reported content of the article.

According to the article, Hubinger, whom it identifies as Anthropic's alignment lead, said the company is "trying its best." He is quoted as saying that Anthropic does not yet have a plan to solve alignment for superintelligence. He is also quoted as saying that Anthropic is not "clearly on track" to develop such a plan. The article's headline characterises the overall warning as saying that AI could "kill all humans."

The report does not state when Coxon's resignation takes effect or what he plans to do next, and it gives no dates for either statement. It mentions no specific Anthropic systems and names no other companies. Coxon's resignation is presented as motivated by concern that companies are moving towards superintelligence without adequate safeguards. Hubinger is presented as acknowledging that Anthropic does not currently have a solution to alignment. The reported disagreement is about whether the pace of AI capability development can be reconciled with the current state of alignment research, and the article does not include further details about either person's role or the company's research roadmap.