Command Palette

Search for a command to run...

Discover

Trump Cancels AI Safety Executive Order Citing Innovation and China Risks

Trump Cancels AI Safety Executive Order Citing Innovation and China Risks
Click to expand

President Donald Trump abruptly called off the signing of a high‑profile executive order on artificial intelligence and cybersecurity on Thursday, citing fears it could slow U.S. innovation and hand China an edge in the AI race, just hours after invitations had gone out to leading tech executives for a White House ceremony.nbcnews +1 The draft order would have created a new framework for the government to review powerful AI models before their public release and bolster cyber defenses across critical infrastructure.apnews +1

The about‑face followed intense, last‑minute lobbying from venture capitalist and Trump adviser David Sacks and other industry figures, who warned that even a nominally “voluntary” vetting process for so‑called frontier AI systems could evolve into a de facto regulatory choke point.nbcnews +1 Trump, who has built his AI policy around deregulation and pre‑empting stricter state rules, told reporters he postponed the order because he “didn’t like what I was seeing” and did not want to “get in the way” of U.S. leadership in the technology.politico +1

Inside the Collapse of a Planned AI Safety Framework

The shelved order was the administration’s most serious move yet toward structured federal oversight of cutting‑edge AI, a notable shift after Trump rescinded President Joe Biden’s sweeping 2023 AI safety order and replaced it with directives focused on “removing barriers” to innovation.ibtimes +1 Drafts described a two‑track approach: government‑led initiatives to harden federal and critical‑infrastructure networks against AI‑enabled cyberattacks, and a voluntary process for companies to provide pre‑release access to their most advanced models for government testing — initially for up to 90 days before launch, a window industry negotiators had pushed to shrink.apnews +1

National security officials argued the move was urgent after Anthropic’s experimental Mythos system reportedly demonstrated the ability to autonomously uncover thousands of serious software vulnerabilities, intensifying fears that frontier AI could supercharge hacking tools.apnews +1 The Commerce Department’s AI testing center had already begun quiet collaborations with major labs, but the new order would have formalized those arrangements and assigned agencies such as CISA and the intelligence community a bigger role in evaluating risks from “covered” models.apnews +1 Critics inside and outside government questioned why the Treasury Department appeared to be given an unusually central role in cyber‑vulnerability coordination, a choice some saw as driven more by internal politics than technical expertise.nbcnews

Innovation vs. Oversight: A Stark Contrast With Biden’s AI Order

Trump’s hesitation underscored a sharp philosophical divide with the Biden‑era AI regime he dismantled. Biden’s October 2023 executive order mandated a broad, cross‑agency program for AI safety, including binding reporting requirements for companies training models above defined compute thresholds, federal red‑teaming of high‑risk systems, and timelines as short as 90–270 days for agencies to develop standards and testing infrastructure.washingtonpost Trump’s 2025 “Removing Barriers” order revoked those obligations, ordered an “AI Action Plan” centered on competitiveness, and was followed by a December 2025 directive instructing federal agencies to challenge state AI rules in court and condition funding on alignment with a national pro‑innovation framework.techbuzz +1

Civil‑rights advocates, worker groups and many AI policy researchers have argued that this approach leaves gaps around bias, consumer protection and safety, especially as Trump’s pre‑emption order moves to undercut stricter state laws in places like California and Illinois.cbsnews +1 Some experts supported the now‑postponed AI and cybersecurity order as a modest step back toward a Biden‑style model‑testing regime, even if formally voluntary; others warned that without clear authorities and strong technical capacity inside government, any new framework risked being either toothless or politicized.cryptobriefing +1 Within industry, companies such as OpenAI have publicly endorsed structured collaboration on safety, while influential figures like Sacks counter that “having the federal government review models before their public release would slow down innovation and harm the U.S. in its AI race with China,” according to one account of his call with Trump.cryptobriefing

The Bigger Picture

The collapse of the AI order showcased the unresolved tension at the heart of U.S. AI policy: how to confront rapidly escalating security risks from frontier models without imposing guardrails that major players can portray as industrial handcuffs. Trump’s decision pleased allies who see any federal review of models as the thin end of a regulatory wedge, but it also left national‑security officials and many researchers without the clearer framework they had sought after high‑profile demonstrations of AI‑driven hacking capabilities.nbcnews +2 With states moving ahead on their own rules, foreign governments tightening oversight, and industry racing to deploy more powerful systems, the question now is not whether the United States will regulate AI — but whether a fragmented, stop‑start approach can keep pace with the technology it is trying to control.