We see anxiety about artificial intelligence in memes and in the midterm election campaigns, but the alarm strikes a deeper chord when the fear comes from within – that is, from those who know the industry. Jacob Coxon, a 27-year-old employee at Anthropic and a researcher who previously worked at OpenAI, has resigned because, he says, Dario Amodei’s company and its competitors are building something they will not be able to control. Coxon, whose role involved feeding the machines as much data as possible, stated that he no longer wished to be part of this race between companies that could spiral out of control and destroy humanity. “As early as next year, things could be out of control,” he said, writing on X that no company is behaving “responsibly”. His public statement sparked reactions from other industry workers, such as a team leader at Anthropic who wrote: “We honestly believe that AI could wipe out the entire human race”, with a risk of over 10 per cent that this could happen within the decade, given that there is no plan for how to manage the “superintelligence” being developed with such fervour in Silicon Valley. At this stage, “extreme caution” is needed, said a scientist at OpenAI, who is concerned about “the consequences for which we are unprepared in the rapid and ongoing development of machine learning”.The race to advance AI is being driven by a desire to counter Chinese efforts, with the aim of being the first to achieve a functional superintelligence – a race that, for many, bears a resemblance to the Cold War-era nuclear arms race between the United States and the Soviet Union. Coxon almost joked about this, saying that whilst the Manhattan Project was once carried out in a secret bunker in the desert, the race for AI is taking place “on the MacBooks of a few engineers living in San Francisco” – something “absurd”. These apocalyptic warnings have also sparked opposing responses, such as that of tech journalist Taylor Lorenz, who criticised the “moralistic doomsaying” that “sows terror amongst the public about a new technology – terror that will lead to the worst possible laws being passed on the subject – and all for what? For a few likes on Twitter?”.In the United States, legislation on AI has stalled, partly because the Trump Administration wants to maximise the economic benefits of AI, as reported by the Wall Street Journal. Not to mention the fact that the President, judging by his social media posts, is fond of images generated by artificial intelligence. Legislative proposals, on the other hand, are coming from some Democrats who want to pause the development of AI, at least until new rules are decided upon. The idea comes from Amodei himself, the CEO and founder of Anthropic, who has called for the process to be slowed down until there are stronger safety measures that keep pace with technological developments. But, Amodei says, he cannot do this alone. The pause must apply to everyone; otherwise, it won’t work. Competitors such as Meta, OpenAI or Chinese companies would press ahead, filling the void: a classic prisoner’s dilemma.