The Trump administration is telling AI companies that any slowdown in frontier development is their responsibility, not a reason for government to suspend competition rules or coordinate the industry. The response, reported by WIRED, adds a substantive government position to a debate previously driven by laboratory leaders and safety researchers.
Anthropic chief executive Dario Amodei had asked for public support to pace development. OpenAI chief executive Sam Altman, Elon Musk and Microsoft chief executive Satya Nadella expressed varying support for caution. The emerging divide is not simply over risk; it concerns who has authority to impose restraint.
The White House Rejects a Need for Permission
David Sacks, co-chair of the President’s Council of Advisors on Science and Technology, wrote that laboratories worried about unreleased models could slow down voluntarily. He rejected the suggestion that companies needed government permission and questioned whether the proposed coordination would amount to suspending antitrust law for a cartel.
Amodei had asked the government to help competing AI companies discuss common standards without breaching competition rules. His options included government participation or a waiver covering specified discussions. He also sought support for international coordination, including with China.
Sacks challenged both premises. He suggested that motivations for slowing development might not be purely altruistic and said China was unlikely to enter a global agreement. Those statements are the administration adviser’s position, not a legal determination that every form of industry coordination would violate antitrust law.
The exchange places laboratory self-interest inside the safety debate. A shared pause might reduce risk, but it could also benefit established companies by limiting challengers. WIRED does not provide evidence that either motive explains Amodei’s proposal; it reports Sacks’s skepticism.
Competition With China Defines the Response
President Donald Trump emphasized US leadership while speaking during a visit to Ireland. He said negative forces were raising concerns that would not occur and argued that whoever wins AI wins more broadly. His comments favor continued development while leaving open a general possibility of guardrails.
House Speaker Mike Johnson made a related argument on CNN. He warned that an emergency congressional attempt to regulate AI could cause the United States to lose its race with China and called for balance rather than rapid intervention.
WIRED says the US government largely treats China’s AI ecosystem as an economic and national-security threat. That framing makes a coordinated slowdown politically difficult even when American executives endorse one. The article does not quantify the countries’ relative capabilities or establish that a particular safety rule would change the competitive balance.
The new development is therefore institutional. Company leaders are asking government to create space for collective restraint, while senior political figures are pushing responsibility back onto individual firms and treating national competition as the limiting condition.
Voluntary Oversight Becomes the Immediate Test
Without government coordination, laboratories are turning to voluntary measures. Altman said OpenAI would support Amodei’s proposal to place independent third-party evaluators inside frontier companies with access comparable to employees. Their role would be to examine safety practices and products and report incidents.
Altman acknowledged that the arrangement would carry significant costs and argued that competitive pressure should not justify recklessness. The source does not identify evaluators, start dates, publication rights or enforcement if a company rejects their findings.
Voluntary access can create scrutiny sooner than legislation, but it depends on permissions granted by the organizations being reviewed. It also does not resolve Amodei’s request for common standards or international agreement.
The White House response clarifies where the burden sits for now. Laboratories that believe the frontier is moving too quickly have been told to change their own release decisions. Whether separate company commitments produce comparable controls—or merely different definitions of responsible pacing—will be the practical test of that approach.
