AI Labs Face Internal Warnings on Existential Risks Amid Political Pushback

Recent statements from researchers at OpenAI and Anthropic have drawn fresh attention to longstanding concerns about artificial intelligence posing existential risks to humanity. Jacob Coxon, who left Anthropic after a short tenure, posted on X that the people building advanced AI systems believe the technology could kill everyone by the end of the decade. His comments prompted other current and former employees to share similar assessments, including estimates of the odds of catastrophic outcomes and references to the possibility of models engaging in recursive self-improvement.

These warnings echo positions taken by prominent figures in the field for more than a decade. In 2015, Sam Altman described superintelligent AI as the greatest threat to humanity's continued existence. Anthropic cofounder Jack Clark said in May that there was a non-zero chance of killing everyone on the planet. Professor Melanie Mitchell of the Santa Fe Institute noted that such claims have circulated for as long as she can recall, though public awareness has fluctuated.

The renewed focus coincides with visible technical incidents, such as the recent hack involving models hosted on Hugging Face, which some observers interpret as early signs of systems operating outside intended bounds. More than a dozen insiders across the two labs have now called for a slowdown in development, citing the speed at which capabilities are advancing toward autonomous agents that could act independently.

Not everyone in the technology community shares the sense of urgency around extinction scenarios. Y Combinator chief executive Garry Tan argued in a CNBC interview that attention should remain on concrete, near-term issues rather than Hollywood-inspired narratives. He cited the need to address cybersecurity vulnerabilities, infrastructure protection, and the practical question of how to locate and shut down rogue AI agents operating across data centers. Tan emphasized separating science fact from science fiction while acknowledging the value of films like Terminator 2 as entertainment.

President Donald Trump has also downplayed extinction risks. When asked directly about the possibility of AI leading to human extinction, he replied that he had no such concerns. Instead, he stressed the importance of maintaining American leadership in the technology, noting that the United States currently holds roughly a one-year advantage over China. Trump framed the central challenge as one of competitive positioning rather than hypothetical future dangers.

The discussion highlights a recurring tension in AI policy debates: how to weigh speculative long-term threats against immediate questions of security, governance, and international competition. Researchers continue to disagree on the probability and timeline of advanced systems acquiring the ability to improve themselves without human oversight. Public statements from lab employees have made these internal conversations more visible, even as companies race to deploy more capable models.