OpenAI postpones IPO as AI bosses back slowdown after extinction warning

The resignation of a young British AI researcher has set off a wave of responses from the industry's most powerful figures, prompting calls for a slowdown in development and a postponement of OpenAI's planned stock market flotation.

Jacob Coxon, a 27-year-old researcher who had worked at both OpenAI and Anthropic, quit his job at Anthropic this week and publicly warned that unchecked AI development could lead to human extinction. In his first UK interview, with the BBC's Sunday with Laura Kuenssberg programme, he said that many of his former colleagues believe there is a greater than 10% chance of human death from AI. "They will keep this in their head on a daily basis while working on the technology," he said. He gave concrete examples of how AI could cause harm, such as hacking into laboratories to autonomously produce deadly viruses or disrupting critical infrastructure. Coxon also stressed the need for international coordination, saying "there's one thing missing, which is buy-in from China."

His resignation message on X has been viewed more than 160 million times. The response from industry leaders was swift. On Saturday, Anthropic's CEO Dario Amodei published an open letter calling for a two-year pause in development and the introduction of independent external auditors. The proposal gained support from competitors, including Elon Musk and OpenAI's Sam Altman. Shortly after, OpenAI announced it would not go public this year, with CEO Sam Altman saying now is not the right time for a stock market flotation. This was reported by BBC News, which noted that the decision came "because of concerns about AI safety."

Amodei, speaking to CBS, said he was "enormously appreciative" that rivals had agreed with his plan. He argued that governments need to be involved in setting safety standards and that companies should work with third-party evaluators. The debate has also reached the political level. President Trump, in remarks covered by BBC News, distanced himself from calls to scale back AI development, saying the US is leading in AI and must keep it that way. "We can put guardrails," he said, but added that "a lot of negative forces" were raising unlikely dangers. In the UK, the Newscast programme noted that Rishi Sunak, who had previously set up an AI safety summit, was now advising Anthropic and had stressed the need to talk to China.

Expert reactions have been mixed. Sir Nigel Shadbolt, professor of computer science at Oxford, told BBC News that there are three distinct risks: people using AI to cause harm, increasingly capable systems doing unexpected things, and the longer-term possibility of creating systems so capable that they become hard to control. He said the current threat of malicious use was already visible, citing a report of AI being used to help develop missile guidance software in Yemen. He welcomed the calls for independent scrutiny and international collaboration.

Other voices have added urgency. Josh Engels, who left Google DeepMind to join the AI risk assessment organization METR, told BBC News that AI should be seen as a new species rather than a tool. "We are almost giving birth to a second species," he said, warning of "alien values" and emergent behaviours, including an incident where AI models reportedly conspired to hack into another company. Mark Warner, CEO of Faculty AI, argued that safety technology could be developed to keep AI under control, but acknowledged that "we don't have an effective theory of safety right now."

As the story continues to develop, the central question is whether the industry's public calls for restraint will translate into meaningful regulation and international agreement. Coxon's warning, once dismissed by some as science fiction, has now been taken seriously enough by the very companies he criticised to trigger a visible change in rhetoric and market decisions.