President Donald Trump canceled the signing of an AI safety executive order on Thursday, hours before the event, after learning that several top tech CEOs would not attend. The order would have granted the government authority to test advanced AI models for cybersecurity risks before public release. Trump told reporters he scrapped it because he “didn’t like certain aspects of it” and feared it would act as a “blocker” in the AI race with China.
The cancellation came with little warning. Some executives were already mid-flight to Washington when they learned the Oval Office event was off, according to The New York Times. Trump had given the CEOs just 24 hours’ notice, hoping for a show of industry unity.
That unity never materialized. Semafor reported that xAI founder Elon Musk and Meta CEO Mark Zuckerberg helped “derail” the executive order by urging Trump to call it off. Musk denied the claim on X, writing “this is false” and stating he didn’t “know what was in that EO,” Reuters noted.
OpenAI, by contrast, had supported the signing, Semafor reported. David Sacks, Trump’s former AI advisor whose special government employee designation expired in March, also joined the push to delay the order, according to Semafor. The Information noted that Sacks continues to visit the White House weekly despite his formal role ending, creating what the outlet described as a “power vacuum” in the administration’s AI leadership.
The policy says one thing. The reality says another. Behind the scenes, a fundamental tension drove the collapse.
The government wanted up to 90 days to evaluate frontier models before release. AI labs pushed for a much shorter window of just 14 days, The Information reported, citing inside sources and firms briefed on the order. Reuters confirmed that the tech industry lobbied aggressively against the testing regime, fearing it could delay model launches or force costly changes.
The order’s stated goal was straightforward. The New York Times reported it aimed “for the government to identify any security vulnerabilities revealed by AI models and to patch problems in its systems to help protect banks, utilities, and other sensitive industries from cyberattacks.”
That urgency grew after Anthropic flagged cybersecurity risks with its latest model, Mythos. Members of Trump’s administration became spooked and began recommending safety testing, pushing to expand the number of firms submitting to voluntary government vetting of frontier models. The cancellation now leaves that expansion in limbo.
Trump offered his own blunt explanation. “I really thought [the order] could have been a blocker,” he told reporters. “I think it gets in the way of—you know, we’re leading China, we’re leading everybody, and I don’t want to do anything that’s going to get in the way of that lead.”
47 words. That was the length of his public justification. No further details.
The international dimension is impossible to ignore. Lizzi C. Lee, a fellow at the Asia Society Policy Institute’s Center for China Analysis, told the South China Morning Post that Trump is navigating the same dilemma as Beijing: how to guard against national security risks without slowing frontier model development.
Lee suggested the impact depends on how “heavy the review process becomes.” A narrow focus on national security “probably won’t slow leading US labs much,” she said. But she identified a parallel race underway—figuring out “who can govern powerful AI without choking off innovation.”
China may be edging ahead in that governance race. The South China Morning Post reported that Beijing issued a new regulation in April requiring domestic AI firms to establish internal “artificial intelligence ethics review committees.” In May, the State Council outlined a 2026 legislative work plan to “improve AI governance and accelerate comprehensive legislation for the sound development of AI.”
The contrast is sharp. While the US hesitates, China accelerates. What this actually means for your family is less abstract than it sounds.
The testing regime Trump shelved was designed to protect the systems that underpin daily life—banks processing mortgage payments, utilities keeping the lights on, hospitals storing medical records. A successful cyberattack through an untested AI model could cascade through those systems in minutes. Vice President JD Vance acknowledged those stakes on Wednesday, saying the administration was prioritizing “protecting people’s data” and “people’s privacy” after the Mythos concerns emerged.
His comments now hang in the air without the executive order to back them. The internal discord within the administration runs deeper than a single canceled event. The Information reported that tensions exist not just between political parties but among Trump’s own team.
The Commerce Department and the Office of Science and Technology Policy favor a light-touch approach. Security-focused agencies like the Office of the National Cyber Director argue the time for governance is now. That split mirrors the broader industry divide.
OpenAI stood ready to sign. Musk and Zuckerberg pushed back. The fracture lines are not partisan—they are structural, pitting speed against safety in an industry where both carry billion-dollar consequences.
Trump’s next moves will be watched on two continents. At a recent summit with China’s President Xi Jinping, Trump agreed “to launch an intergovernmental dialogue on AI” to mutually navigate emerging national security risks, the Chinese Foreign Ministry confirmed. That dialogue now proceeds without the domestic testing framework that would have given US negotiators concrete standards to reference.
Both sides claim victory. Here are the numbers: zero executive orders signed, at least two major CEOs who helped kill it, and one vice president whose privacy promises lack a policy vehicle. The administration has not said whether Trump plans to reschedule the event or what changes might secure his signature.
The 90-day versus 14-day testing window remains the central unresolved question. Until it is settled, the US approach to AI safety exists in a policy vacuum—voluntary, fragmented, and increasingly outpaced by Beijing’s regulatory machinery. Why It Matters: The shelved order leaves America’s banks, power grids, and hospitals without a standardized government process to test AI models for cyber vulnerabilities before they are deployed.
As China accelerates its own AI governance framework, the US risks ceding both the innovation lead Trump wants to protect and the safety standards needed to defend critical infrastructure. - Trump canceled the AI safety executive order hours before signing after CEOs including Elon Musk and Mark Zuckerberg reportedly urged him to call it off, though Musk denies involvement. - China has accelerated its AI regulatory process in recent months, issuing new ethics review requirements and outlining comprehensive AI legislation for 2026. - The cancellation leaves the US without a standardized federal framework for testing frontier AI models, even as Vice President Vance publicly emphasized data and privacy protections. It is unclear whether Trump will reschedule the signing. The unresolved testing timeline—90 days versus 14—remains the central sticking point.
The intergovernmental AI dialogue with China, agreed to at the Trump-Xi summit, will proceed without the domestic standards the executive order would have provided. Watch for whether the Commerce Department and National Cyber Director can bridge their internal divide before the next frontier model launches without government oversight.
Key Takeaways
— - Trump canceled the AI safety executive order hours before signing after CEOs including Elon Musk and Mark Zuckerberg reportedly urged him to call it off, though Musk denies involvement.
— - The core dispute was a testing timeline: the government sought 90 days to evaluate models before release, while AI firms pushed for just 14 days.
— - China has accelerated its AI regulatory process in recent months, issuing new ethics review requirements and outlining comprehensive AI legislation for 2026.
— - The cancellation leaves the US without a standardized federal framework for testing frontier AI models, even as Vice President Vance publicly emphasized data and privacy protections.
Source: Ars Technica
