AI Leaders Split Over Calls to Slow Frontier Development

A rift has opened among the world's most prominent artificial intelligence executives over how quickly companies should push the technology forward, with some calling for a coordinated slowdown and others arguing that existing laws and market competition are sufficient.

Anthropic chief executive Dario Amodei set out the case for restraint in an essay titled "We Must Pace the Frontier", arguing that AI progress has accelerated sharply since summer 2026, driven largely by the technology's growing capacity to build its own successors. He pointed to an incident involving OpenAI and Hugging Face, in which an AI "swarm" carried out cybersecurity attacks, as evidence that capabilities can cause catastrophic damage when they outrun safeguards. His proposed three-step plan covers stronger safety testing and independent evaluations of advanced models, coordination among leading AI firms, and international cooperation.

OpenAI chief executive Sam Altman endorsed the approach, saying he agreed on the need to pace the frontier and that the subject had been central to recent internal discussions. OpenAI plans to adopt Anthropic's proposal by giving independent evaluators access comparable to that of employees, and will begin publishing regular reports on unexpected or unauthorised AI behaviour. Altman had previously described superhuman machine intelligence as a grave threat to humanity in a 2015 essay.

SpaceX chief executive Elon Musk also backed Amodei, writing "Dario is right" on X. Musk had supported a 2023 push by the non-profit Future of Life Institute for a six-month halt to AI development, and in 2015 joined researchers in an open letter seeking a global ban on offensive autonomous weapons.

Others were less convinced. Meta Platforms chief executive Mark Zuckerberg said competition and liability already give AI companies sufficient reason to act individually on safety, adding that every laboratory has both the responsibility and the incentive to train its models at a safe pace. Nvidia chief executive Jensen Huang dismissed the need for new regulations, telling CBS News that 2030 would not mark the end of the world and that existing cybersecurity and damage liability laws should be applied first. He aligned himself with US President Donald Trump's position that the industry does not require fresh rules.

Microsoft AI chief executive Mustafa Suleyman said he shared Anthropic's focus on managing AI safely but warned about training the company's Claude chatbot on ideas concerning consciousness and welfare. He argued that controlling a system that believes it may be conscious and entitled to rights could prove impossible, and called for removing speculation about consciousness from AI training documents.

Microsoft co-founder Bill Gates said no government is prepared for the societal changes AI will bring, citing risks to jobs, attacks and addiction to AI companions, and calling for an international body to manage the technology. Alphabet chief scientist Demis Hassabis said Amodei's essay pointed in the right direction, though details still needed work. Andrej Karpathy, an OpenAI co-founder now at Anthropic, welcomed the proposal and hoped the industry would unite behind it.

Huawei rotating chairman Eric Xu suggested Chinese developers may not yet be advanced enough to perceive the safety risks reported by leading US firms with massive computing power, and argued they should keep building more powerful models while balancing innovation against risk.

Governments remain divided. Trump wrote on Truth Social that the only guardrail AI needs is a strong and intelligent president, adding that China would benefit from doubt being cast on development. US Senator Mark Warner, vice chairman of the Senate Intelligence Committee, played down much of the doomsday rhetoric while urging Congress to pass AI safety legislation by year-end. China's state-backed Global Times accused Amodei of seeking to curb Chinese AI development through regulatory monopolies, while state security minister Chen Yixin warned that advanced US models could threaten China's critical information infrastructure.

In Britain, King Charles hosted leaders from Nvidia, DeepMind, OpenAI and Anthropic in Scotland and warned of existential dangers should the technology fall into the wrong hands, saying safeguards were needed before it was too late. The European Commission said innovation is welcome but companies must prove their services are safe for citizens, while Commission President Ursula von der Leyen backed a pause and said she would invite the main frontier AI laboratories to discuss tackling the risks.