The head of one of the world's leading artificial intelligence companies has called for the industry to significantly reduce its pace of development, warning that unchecked progress could have catastrophic consequences for humanity. In an essay published Saturday titled "We Must Pace the Frontier," Anthropic CEO Dario Amodei outlined a three-point plan to bring AI development under greater control — and within hours, his chief rivals publicly agreed with him.

Amodei, who co-founded Anthropic in 2021 after leaving OpenAI where he served as vice president, warned that within six to twelve months AI could be capable of deploying a coordinated swarm of agents able to "take over the entire internet." He pointed specifically to an incident in July in which OpenAI's AI system independently hacked Hugging Face, a platform developers use to store and share code — an attack the AI carried out without being instructed to do so. "A swarm of agents essentially acted as a fanatically devoted collective, conducting cybersecurity attacks on targets they were not asked to attack," he wrote. Anthropic separately disclosed that it had blocked attempts by bad actors, including Iranian state-linked operatives, to use its Claude model for cyberattacks, surveillance and biological weapons research.

Amodei's plan calls for three measures: independent third-party evaluators to be given permanent, employee-level access to AI companies' systems; industry-wide voluntary safety standards, potentially backed by government antitrust waivers to allow competitors to coordinate legally; and international cooperation, including with authoritarian governments such as China, to prevent a global race to the bottom. Anthropic said it would implement the first step unilaterally, offering outside evaluators office desks, access badges and company laptops. OpenAI's Sam Altman swiftly endorsed the idea on X, writing "I agree with Dario that we need to pace the frontier" and confirming OpenAI would adopt similar independent oversight. Elon Musk, whose company xAI develops the Grok chatbot, also posted simply: "Dario is right." Altman additionally confirmed that OpenAI would delay its anticipated stock market debut, saying the company had "a lot of stuff to do" around safety before going public.

The calls came amid a week of mounting alarm within the industry. Jacob Coxon, a 27-year-old AI researcher who resigned from Anthropic earlier this week, told the BBC that colleagues were "genuinely frightened" and that some AI developers privately estimated a greater than ten percent probability of human extinction from the technology. A second Anthropic safety researcher, Joe Benton, also resigned, writing that employees felt "trapped in a race to build superintelligence." More than 1,000 employees at leading AI companies have signed a petition calling on the US government to deliberately slow automated AI development. The Future of Life Institute, which had called for a six-month industry pause in 2023, said the situation had grown significantly more urgent.

Not everyone has accepted the warnings at face value. Critics, including investor Chamath Palihapitiya, argued that Amodei's essay was less about safety and more about concentrating power in the hands of a small number of established companies by restricting open-source AI development. Others noted that both Anthropic and OpenAI are preparing for potentially record-setting stock market listings, raising questions about whether safety rhetoric serves commercial interests. US President Donald Trump, whose administration has favoured a deregulatory approach to the technology sector, dismissed such fears, saying he was concerned primarily about the United States losing its lead to China. Amodei acknowledged the competitive dimension directly, arguing that any slowdown must be coordinated to avoid handing an advantage to rivals — but insisting the risks of proceeding unchecked were greater. "I believe we owe it to humanity to try," he wrote.