• Early plans for a frontier‑AI rulebook
• Rising concern after cyber‑capable models
• Tech pushback and Trump’s skepticism
• The last‑minute cancellation
• Aftermath: uncertainty over AI safety rules

Trump Cancels Signing of AI Safety Executive Order President Donald Trump’s last‑minute decision to halt an AI safety executive order has exposed deep divisions over how — and whether — Washington should oversee powerful new AI models.

§Early plans for a frontier‑AI rulebook

In mid‑May, the White House was preparing a sweeping order on cybersecurity and “covered frontier models,” aiming to give the government early access to advanced AI systems and to strengthen protections for national security agencies and critical infrastructure. The draft envisioned a “voluntary framework” under which AI labs would share models with the government up to 90 days before public release. A separate draft obtained by POLITICO showed similar plans for voluntary pre‑release reviews, explicitly stressing that “nothing” in the order authorized mandatory licensing or preclearance of new AI models.

Supporters inside and outside government framed this as a first step to avoid a catastrophic “Chernobyl moment” in AI, arguing that structured testing of frontier models is needed before they can be widely deployed.

§Rising concern after cyber‑capable models

Officials were jolted by systems like Anthropic’s Mythos and OpenAI’s GPT‑5.5 Cyber, which can “quickly find and exploit security vulnerabilities,” prompting calls for a formal pre‑release evaluation process led by the Office of the National Cyber Director and other agencies. The goal, according to reporting on the now‑stalled order, was for government testing to identify serious cyber risks before models reached the public.

§Tech pushback and Trump’s skepticism

As a signing ceremony was planned for May 21, Axios reported that the order had become a focal point of infighting within the administration and the tech sector. Trump adviser and former “AI czar” David Sacks “hated it,” and a source said the whole effort was “just something doomers wanted.”

Major industry figures including Meta CEO Mark Zuckerberg and xAI CEO Elon Musk discussed the order directly with Trump in the hours before the event, with reporting indicating they helped “derail” the plan by urging him to call it off. (Musk has publicly denied having a hand, saying he didn’t “know what was in that EO.”)

At the same time, many tech firms worried that government safety testing could delay model launches or force design changes. One key dispute was timing: officials sought up to 90 days of review, while labs pushed for as little as 14 days.

§The last‑minute cancellation

Hours before the scheduled Oval Office photo op with AI CEOs, the event was abruptly pulled after some top executives declined invitations sent with only 24 hours’ notice, reportedly infuriating Trump. Several executives were already “midair on their way to the Oval Office” when they learned the signing was off.

Trump told reporters, “I didn’t like certain aspects of it. I postponed it,” adding that the order “could have been a blocker” for jobs and the “tremendous good” AI is creating. He repeatedly cited China, saying, “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.” He has long “just hate[d] regulation,” one source said.

§Aftermath: uncertainty over AI safety rules

With the signing shelved, Axios reported that “the industry and administration are scrambling to figure out what’s next” as key questions over government access to top models and AI safety remain unanswered. For now, accelerationists favoring light‑touch rules “have won out,” while advocates of stricter safeguards warn that delaying the order postpones what they see as a necessary first move to prevent a future AI disaster.

Continue reading https://foxvector.com