AI Labs Face Internal Warnings on Existential Risks Amid Political Pushback

Former Anthropic employee Jacob Coxon quit the company this week and warned that researchers at the top AI labs believe their own technology could wipe out humanity by 2030. His post triggered a wave of similar claims from other workers at OpenAI and Anthropic, with some estimating the odds of catastrophe in the coming years. The sudden public debate has produced plenty of headlines about killer machines, yet President Trump made clear he sees no such threat.

Trump told reporters he has no concerns about AI leading to human extinction. Instead he stressed the need to maintain America's lead over China, which he described as roughly a year behind. The president said falling behind would put the country in a bad position, framing the real stakes around national strength rather than speculative end-of-the-world scenarios.

Y Combinator CEO Garry Tan echoed that view in a CNBC interview. He called for separating science fact from science fiction and pointed to practical dangers like cybersecurity and infrastructure security instead of Hollywood-style robot uprisings. Tan noted that questions about shutting down rogue AI agents or tracing them to specific data centers matter far more than debates over recursive self-improvement turning machines into superintelligent overlords.

The warnings from lab insiders are not entirely new. Sam Altman raised similar alarms about superintelligence as early as 2015, and Anthropic co-founder Jack Clark has spoken of a non-zero chance of killing everyone on the planet. Professor Melanie Mitchell observed that such claims have circulated for as long as she can remember. What changed this time appears to be the blunt language from recent departures and the timing amid broader public unease.

Coxon described the situation as the labs gambling with lives. Other researchers have highlighted risks from models that could improve themselves without human oversight. Yet these same labs continue to race ahead with bigger models and faster releases. The pattern suggests the doomsday talk may serve to shape policy debates or slow competitors while the core work proceeds.

Trump's focus on the China competition aligns with the immediate reality of rival nations developing advanced systems. Losing that edge carries concrete consequences for defense, economy, and technological independence. Fears of total extinction, by contrast, remain unproven speculation that distracts from measurable problems such as data center vulnerabilities and coordinated attacks on critical systems.

The episode shows how quickly Silicon Valley insiders can amplify dramatic narratives. Public attention now turns on whether those warnings reflect genuine caution or a bid to influence regulation and funding at a critical moment.