OpenAI's Sam Altman has spoken at length about the potential hazards of fast-moving artificial intelligence development, adding his voice to that of Anthropic's Dario Amodei in urging the technology sector to moderate progress on its most advanced systems.

In a series of posts on X just after midnight on September 14, Altman outlined two potential negative outcomes for AI development: a scenario in which humanity loses oversight of the future to AI systems, and one in which a single individual or entity accumulates disproportionate power. Altman expressed support for a federal framework establishing uniform safety standards for frontier AI, and stated that competitive pressures should not be used to justify inadequate safeguards.

He also clarified that moderating the pace of AI development is distinct from halting it altogether, noting that while progress has been swift and will continue, it would be preferable for it to advance more slowly than it otherwise might, given that measures such as safety assessments and monitoring carry considerable costs.

These remarks follow a proposal published on September 12 by Amodei, who contended that recent advances in the field have begun to exceed the industry's capacity to comprehend and govern increasingly capable systems. Amodei outlined a three-part plan intended to moderate the rate of advanced AI progress while preserving the competitive standing of domestic technology firms and preventing overseas rivals from gaining ground.

Under the initial measure, independent safety assessors at advanced AI laboratories would receive access comparable to that granted to employees. Anthropic has pledged to take that step. Amodei also urged leading AI firms to develop shared safety frameworks and, over time, pursue international coordination — including with China — to ensure that no company or nation faces a competitive penalty for proceeding more cautiously.

OpenAI's chief executive indicated the company would extend the same level of access to independent evaluators as Anthropic had pledged to provide. Elon Musk also expressed support for Amodei's position, while Demis Hassabis of Google DeepMind described the proposal as pointing in the right direction, though any coordinated deceleration would need to contend with the ongoing technology competition between the United States and China.

In his proposal, Amodei addressed that tension directly. He contended that Beijing's involvement would ultimately be necessary for any effective global framework, yet simultaneously urged the United States to uphold export controls on advanced artificial intelligence chips and semiconductor manufacturing equipment destined for China.

Trump dismissed calls to decelerate artificial intelligence development in the United States, asserting that the country currently holds an advantage over China in the field. Trump told reporters during a visit to Ireland on September 13 that the United States held a position of global leadership that he intended to preserve, adding that dominance in artificial intelligence would be decisive and that regulatory guardrails remained an option.

The quoted remarks were followed on September 14 by a post on Truth Social in which the president described what he called a coordinated effort against AI and data centers, while asserting that federal authorities hold the power to regulate and take legal action against AI companies where circumstances require it.

David Sacks, the White House's technology policy adviser, noted in an X post that AI developers such as OpenAI and Anthropic already hold the discretion to voluntarily pause their own research if internal assessments identify potential hazards in unreleased models, and that no outside approval is needed for them to act. Sacks also questioned whether China would participate in any multilateral framework on the issue.

China's Foreign Ministry responded to Amodei's claim that a Chinese lead in artificial intelligence would threaten the United States and the world, with spokesperson Guo Jiakun telling a press briefing on September 14 that cultivating alarm, adversarial positioning, and destructive competition would only hinder efforts toward responsible global AI governance, serving no one's interests. A publication linked to the Communist Party released an editorial characterising Amodei's proposal as an outdated geopolitical strategy aimed at restricting China's advancement in technology.

The exchanges arise ahead of a scheduled September 24 bilateral meeting between Trump and Xi Jinping in Washington, at which artificial intelligence is expected to feature on the agenda.