President Donald Trump abruptly postponed an executive order signing that would have established a federal pre-certification system for the AI industry just before the ceremony.
According to Politico and other outlets, on the 21st (local time) Trump halted the signing, saying, “I didn’t like what I was seeing,” and adding, “The United States leads every country, including China, and we don’t want any action that could jeopardize that lead.”
He said, “AI brings tremendous benefits and creates many jobs,” and warned that “this executive order could become an obstacle.”
The White House had sent invitations to major tech companies and industry groups a day earlier to prepare for the signing. Executives from OpenAI, Google, Anthropic, Meta, and Microsoft were invited; Anthropic CEO Dario Amodei was reportedly among those expected to attend.
But a few hours before the event, the White House suddenly notified guests that the signing was postponed, creating confusion. Some executives were reportedly already en route to Washington.
The executive order would have marked a shift from the Trump administration’s prior pro-AI, low-regulation stance by introducing a government pre-certification regime for advanced AI models. Under the plan, AI firms would voluntarily share advanced models with the government up to 90 days before public release, allowing the Treasury Department, the National Security Agency (NSA), and the White House’s National Cyber Directorate to assess risks.
The move was reportedly prompted in part by an episode involving Anthropic’s “Misos” model. The order focused on using such models to identify security vulnerabilities in government systems and critical infrastructure—banks, hospitals, and power grids—before they could be exploited, and on strengthening defenses. A primary goal was to protect small hospitals and regional financial institutions that lack robust cyber defenses.
But parts of Silicon Valley and some White House officials pushed back, arguing the measures could slow innovation. David Sacks, a venture capitalist and a key AI adviser to President Trump, recently said on a podcast that an FDA-style approval process “could be lethal to innovation.”
When some administration officials floated treating AI models like drugs—requiring safety validation before release—the industry warned that regulators were going too far.
Analysts say the split reflects two camps inside the White House on AI policy: one advocating stronger government oversight to address national security and cyber threats, and the other arguing that minimizing regulation is necessary to maintain the U.S. lead over China.
Since returning to office, President Trump has rolled back many of the Biden administration’s AI regulations. The Biden administration had required firms to notify the government and submit safety-assessment results when developing advanced AI models; the Trump team criticized those rules as overregulation that would stifle business.
Public opinion, however, has shifted as AI has advanced rapidly. A survey by the conservative Institute for Family Studies found that more than 80% of American voters support safety testing of AI models before public release.
Concerns are growing that AI could lead to job losses, a proliferation of large data centers, and new national security risks.
Reporter Chan Park cpark@aitimes.com
Most Commented