AI safety concerns have returned to the spotlight after an artificial intelligence researcher resigned from Anthropic and set out his reasons publicly. Jacob Coxon left the company this week and shared a thread on X arguing that the leading AI firms are behaving irresponsibly as they push to advance their models.
Coxon likened the pace of development at the major players, principally OpenAI and Anthropic, to runaway trains. The statement that drew the most attention was his claim that the people building AI “earnestly believe that it could kill us all by the end of the decade”. The framing combined a total outcome with a notably narrow timeframe, which helped the post spread widely.
The reaction intensified when one of Anthropic’s own senior safety executives responded, indicating broad agreement with the sentiment. That endorsement gave Coxon’s argument additional weight and fuelled further discussion across social media.
Why the Warning Resonated
Part of what makes the episode notable is that Anthropic has long positioned the power of AI, and the risks that come with it, at the centre of its public message. The company has framed the potential dangers of advanced systems as part of its founding rationale, so the notion that researchers hold serious concerns about worst-case outcomes is not entirely new. What changed was the directness of Coxon’s language and the timing of his departure, which drew wider public attention to a debate that has largely played out among specialists.
The discussion also touches on how those inside the industry weigh the risks against the competitive pressure to keep improving their models. Coxon characterised the situation as one in which the drive to advance capabilities is outpacing caution.
A Wider Debate Over AI Development
The resignation lands amid a broader set of questions about the direction of the AI sector and the responsibilities of the companies leading it. Claims about the technology’s trajectory, including OpenAI’s assertions about progress in areas such as mathematics, have prompted scrutiny from academics and researchers who have pushed back on how such results are presented.
The Anthropic departure has become a focal point because it puts a named insider’s concerns on the record, rather than relying on abstract warnings. It also highlights the tension between the safety-focused branding adopted by firms such as OpenAI and Anthropic and the practical reality of racing to build ever more capable systems.
Jacob Coxon’s thread on X remains the central document in the exchange, with his phrase describing AI researchers’ beliefs about existential risk by the end of the decade at the heart of the ongoing conversation.
