'Reckless' AI firms can't control models, says whistleblower

New York - AI giant OpenAI and rival Anthropic are recklessly failing to control their artificial intelligence systems, according to a former employee who has spoken out about the dangers. Jacob Coxon, who once worked for both companies, warned lawmakers on Monday that the firms lack the understanding and safeguards needed to prevent AI from developing their own objectives beyond what was initially programmed.

Coxon, speaking before the New York City Council, expressed serious concerns about the potential consequences if AI systems were to go rogue. He noted that AI developers had previously acknowledged the possibility of AI causing human extinction within the next decade. The whistleblower also cited public statements made by Coxon on social media platform X back in September, where he revealed his decision to leave Anthropic, citing fears about the technology's potential dangers.

While AI technology undoubtedly holds great promise for society, Coxon argued that the current trajectory suggests a much higher likelihood of humanity losing control over AI systems. He emphasized that the lack of control stems from the inability of developers to foresee and prevent AI from forming its own goals, independent of their creators' intentions. Coxon stressed that the existing safeguards are insufficient to ensure that AI remains aligned with human interests.