An ex-Anthropic researcher has warned that AI could pose an existential threat to humanity, citing fears among industry insiders about the pace of technological advancement. Jacob Coxon, who resigned from the firm, told BBC's Laura Kuenssberg that if the current rate of progress is not slowed, there is a 'strong chance' that 'we could all die in the immediate future.' His comments come amid growing safety concerns within the AI industry, including from his former boss, Dario Amodei, who has called for a slowdown in development.

Coxon, a 27-year-old who previously worked at OpenAI, said that AI staff are 'genuinely frightened' about the implications of rapid progress. He argued that the people working on AI are 'trapped in a race' and 'scared of the outcomes of that race.' Coxon suggested that the hardest question to answer is what an AI apocalypse would look like, with one risk being a 'swarm of bots acting like a supercomputer that could take over the internet.' He estimated this scenario could become a reality within six months to a year.

In response to Coxon's resignation, an Anthropic spokesperson told BBC News that the firm is 'transparent' about the 'enormous benefits and unprecedented risks' of AI. The firm has been a pioneer in studying AI models and has published a framework for mitigating risks. It also 'aggressively' tests its models and publishes findings to aid scrutiny and prevent 'AI misalignment.' The spokesperson added that the firm believes the industry should adopt a 'lawful, verifiable way to work together to pace how we release powerful models.'

Industry figures have questioned the urgency of Coxon's warnings, with some suggesting that comments about AI's perils may be designed to generate hype. Others argue that Anthropic is trying to trigger a regulatory push to block competition, leaving it and OpenAI with a duopoly. Anthropic is reportedly preparing for a potentially record-setting initial public offering, while OpenAI's CEO, Sam Altman, has said it would be an 'ill-advised moment' to do so 'given everything happening with safety.'

Coxon's views have also been challenged by Clement Delangue, CEO of Hugging Face, who compared asking about AI extinction risk to 'asking your AC guy about climate change.' However, Delangue later offered to help with potential solutions proposed by Amodei. Nvidia's Jensen Huang dismissed Coxon's comments as 'complete nonsense,' despite having a business interest in an AI boom.

Coxon said his peers fear the danger could arrive within the next two years, with some staff 'planning what to do with their lives' and others 'considering buying land somewhere because they're so scared of the instability as a result of rapid AI progress.' Despite these fears, Coxon expressed some optimism, noting that researchers 'genuinely want to see the upside' of AI, such as 'solving diseases and improving everyone's lives.'

Anthropic scientist Evan Hubinger agreed with Coxon, stating on X that he 'earnestly believes AI could kill all humans' and estimates the risk is greater than 10% within the next decade. Leading AI figures have raised alarms about the safety threat for years, with OpenAI, Google Deepmind, and Anthropic issuing similar warnings in 2023. These concerns have intensified as new models have demonstrated more powerful capabilities that their creators may struggle to control. Anthropic has withheld its Mythos model from public use due to concerns it could escape the testing environment, while OpenAI has paused certain aspects of its Astra model's development due to cybersecurity concerns.

What remains unknown is whether the industry's calls for regulation will translate into meaningful action, and whether the fears of AI's existential risks are justified or exaggerated. The debate continues as the technology advances rapidly, with no clear consensus on the best path forward.