President Donald Trump has signed an executive order aimed at overseeing artificial intelligence (AI). This move comes shortly after a previous postponement due to concerns about its impact on America’s technological advantage.
The order sets a framework for the federal government to evaluate the national security risks posed by advanced AI systems. It allows a review period of up to a month before their public release, although participation by AI developers is voluntary.
According to the order, advanced AI can strengthen the nation but also raises national security issues. It calls for coordinated actions from executive departments and agencies.
The specific differences between this order and a previous version that Trump declined to sign on May 21 remain unclear. Notably, the order limits the review period to 30 days. This shorter timeframe aims to avoid being overly burdensome in a rapidly evolving, competitive industry.
Trump’s cancellation of a previous meeting with tech executives stemmed from dissatisfaction with an earlier draft of the order. He emphasized the importance of maintaining the U.S. lead in technology, particularly in comparison to China.
The current directive involves voluntary cooperation with U.S.-based tech companies. These companies, known as “frontier labs,” are focused on developing the most advanced AI systems. Companies like Anthropic, OpenAI, and Google were part of this collaboration, although the formal signing lacked a ceremonial event.
The White House clarified in a social media post that the order establishes a voluntary process for frontier labs to share advanced cyber models. This initiative aims to secure critical infrastructure and bolster government cyber defenses. It does not encompass oversight of all new models, to avoid stifling innovation and free speech.
Policy analyst Juan Londoño from the Cato Institute described the order as a positive step toward preparing the nation for advanced AI systems. He appreciated the voluntary nature of the process but expressed concerns about the lack of clarity regarding the selection of AI models for scrutiny and the role of the National Security Agency (NSA). Londoño warned that granting extensive discretion to the NSA director sets a dangerous precedent, potentially allowing government exploitation of the policy against competing companies.
This order follows Anthropic’s April announcement of Claude Mythos, its latest AI model. Mythos’ potential to identify cybersecurity vulnerabilities raised alarm, prompting meetings between Treasury Secretary Scott Bessent, Federal Reserve Chair Jerome Powell, and Wall Street CEOs. Anthropic has restricted Mythos’ access to trusted partners but plans to expand this group.
Both Anthropic and its competitor OpenAI view Trump’s order as a significant step in maintaining U.S. leadership in AI. Google also echoed this sentiment. Chris Lehane, OpenAI’s chief global affairs officer, stressed the need for safety frameworks informed by technical expertise to build public trust and accountability.
Democratic Sen. Mark Warner welcomed the order, though he criticized the administration for revoking prior AI guidelines from former President Joe Biden, only to later recognize their necessity.