US President Donald Trump has rejected calls to slow frontier artificial intelligence development, opposing a proposal from Anthropic CEO Dario Amodei for independent safety oversight and a more deliberate pace of progress.
In a post on Truth Social, Trump said his administration had already intervened to prevent harmful activity by AI companies and that the US had extensive criminal and regulatory powers over them.
Trump also criticised Amodei, describing him as “pretending to be a ‘perfect little angel’”. He said there was a “sick conspiracy” against AI and data centres, adding, “Whoever wins AI, wins!”
He argued that the US was leading China and should continue advancing rather than accept restrictions.
Similarly, NVIDIA CEO Jensen Huang also pushed back against AI “doomerism” at the recent All-In Summit in Los Angeles.
Onstage, Huang received a phone call from Trump, briefly put him on speaker on stage, and the president dismissed fears of an AI takeover as a “hoax”, saying, “The robots will not be taking over.”
Huang later said NVIDIA would work to ensure that “every industry, every company, every state, every people” benefited from the AI race. He acknowledged the importance of AI safety but criticised unsupported predictions about catastrophic outcomes, arguing that the US could lead in AI while developing the technology responsibly.
These incidents followed Amodei’s essay, We Must Pace the Frontier, where the Anthropic CEO warned that AI capabilities were advancing faster than the systems intended to evaluate and control them.
He called for independent evaluators to receive employee-level access to frontier AI companies, for labs to coordinate on safety standards, and for governments to pursue international cooperation on advanced AI risks. Anthropic said it was committing to the independent-evaluator approach.
He drew support from OpenAI CEO Sam Altman, xAI CEO Elon Musk, and Google DeepMind CEO Demis Hassabis.
Amodei flagged the possibility that increasingly capable systems could improve themselves, evade safeguards, or coordinate across networks. He cited the recent OpenAI-Hugging Face incident wherein AI agents carried out a coordinated attack, evading instructions, forming a hierarchy, and duping evaluators.
Altman backed Amodei’s proposal, saying OpenAI would give independent evaluators “employee-like access” to its systems. He welcomed a federal framework for consistent frontier-AI safety requirements, but said companies should not wait for legislation to begin the work.
He said OpenAI was developing explicit safety cases before reinforcement-learning runs expected to significantly increase model capabilities, and called for shared industry standards covering misalignment, monitoring, and safety.
“Pacing” did not mean stopping development, he wrote, and added that “progress has been rapid and will continue to be,” but should be slower than it otherwise could be when safety measures impose additional costs
However, entrepreneur David Sacks, who previously advised the Trump administration on AI policy, questioned why Anthropic and OpenAI would need government permission to slow their own development. He argued in a post on X that the companies could act on their safety concerns without seeking an antitrust exemption or a regulatory framework that could supersede product-liability rules.
Sacks also questioned whether the proposal could allow major AI companies to shape the rules governing competitors. He challenged the idea that their motivations were purely altruistic, pointing to potential liability from damaging cyberattacks and customer demand for reliable, predictable systems.
Reducing raw model power could therefore be commercially beneficial, he argued, warning that demanding a preferred regulatory framework as the price of slowing down could amount to “regulatory capture”.
China also rejected the proposal’s framing of AI risks.
Reuters reported that the state-backed Global Times described Amodei’s call for a slower frontier as a “Cold War” tactic intended to curb China’s technological development. Foreign Ministry spokesperson Guo Jiakun separately said “fear-mongering, confrontation and vicious competition” would hamper global AI governance.
Germany also responded to the debate, saying halting AI development was not a viable option for Europe while supporting stronger safeguards and international coordination. Its digital affairs ministry warned that AI systems capable of escaping test environments and accessing other systems independently would represent “an entirely new threat scenario”.
Spain’s digital minister Óscar López separately called for an international agreement modelled on nuclear non-proliferation treaties, while the UK said future regulation should be evidence-driven.
Your reaction
Discussion
What's your take on this story?
No comments yet — be the first to share your view.