On June 2 local time, US President Donald Trump officially signed a much-delayed artificial intelligence executive order. After rounds of bargaining between US federal regulators and Silicon Valley tech giants, the original mandatory 90-day pre-release review was revised into a voluntary 30-day security assessment framework.

Titled Promoting Advanced Artificial Intelligence Innovation and Security, the executive order requires AI developers to voluntarily provide frontier AI models to the US federal government for safety testing up to 30 days ahead of public launch. It explicitly rules out any mandatory pre-approval, licensing or government authorization for AI development and commercial release, leaving enterprises full discretion to opt in or out of the review system. The policy was initially scheduled for signing in late May, yet Trump suspended the signing ceremony abruptly over worries that rigid regulation would hinder America’s AI edge and leave room for China to narrow the tech gap. After closed-door negotiations among the US Treasury Secretary, Defense Secretary and tech representatives, policymakers agreed to slash the review timeline as a compromise.

The regulatory shift was largely triggered by Anthropic’s Mythos AI model, which is capable of detecting security loopholes across healthcare, finance and critical infrastructure systems. The model’s ongoing global commercial trial sparked grave cybersecurity concerns among US national security authorities and accelerated the rollout of the new AI governance rules. Under the new rules, the US Department of the Treasury will set up an AI cybersecurity coordination hub, and Washington will select trusted industry partners to gain early access to cutting-edge AI technologies to reinforce cybersecurity of core public facilities.

While tech firms including Microsoft welcomed the new policy, Democratic US lawmakers criticized the loose regulatory standard for fostering an unregulated “Wild West” environment for AI development. Industry analysts commented the compromise strikes a tentative balance between national security risk prevention and preserving domestic AI innovation vitality, revealing America’s persistent dilemma between AI safety supervision and global industrial competition.