The White House has taken a notable step into the rapidly evolving artificial intelligence debate. President Donald Trump has signed a new executive order aimed at creating a framework for evaluating powerful AI systems before they become widely available, a move that signals a change in the administration’s approach toward regulating the technology sector.
The order arrives at a time when AI companies are racing to release increasingly capable models while governments around the world struggle to determine how much oversight is necessary. Although the new directive is reportedly less aggressive than earlier proposals discussed behind closed doors, it still represents one of the most significant AI-related actions taken by the Trump administration so far.
New Review Process Targets Critical Infrastructure Risks
Under the executive order, federal agencies will work on a system designed to identify software vulnerabilities that advanced AI models may uncover. The goal is to ensure that operators of critical infrastructure—including hospitals, banks, and utility providers—receive information about potential security weaknesses before those vulnerabilities become widely known.
The Office of the National Cyber Director has been tasked with developing the framework. According to reports, some AI companies will be asked to voluntarily provide access to their most advanced models approximately 30 days before public release. Government officials would then evaluate potential cybersecurity risks and coordinate with affected organizations if serious vulnerabilities are discovered.
The final version of the order appears to be a compromise following weeks of discussions between administration officials and technology industry leaders. Earlier proposals reportedly suggested review periods of up to 90 days before public launches, a timeline that sparked concern among companies worried about delays to product releases. The version ultimately signed significantly reduces that review window while keeping the overall framework intact.
The decision follows reports that Trump delayed the original signing ceremony after expressing concerns about portions of the draft order. Behind-the-scenes negotiations eventually resulted in a scaled-back version that administration officials believed would balance security concerns with industry interests.
A Different Direction for Trump’s AI Policy
Perhaps the most surprising aspect of the executive order is what it reveals about the administration’s evolving view of artificial intelligence. Over the past year, the White House generally favored a lighter regulatory approach, often emphasizing innovation and competitiveness rather than oversight.
Much of the administration’s previous AI policy focused on ideological concerns rather than technical risks. Earlier directives sought to prevent federal agencies from using AI systems viewed as promoting political or social agendas. At the same time, the administration challenged efforts by individual states to impose their own AI regulations, arguing that a patchwork of state laws could hinder innovation.
That makes the new cybersecurity-focused order a notable departure. Rather than concentrating on political concerns, the latest directive acknowledges the possibility that powerful AI systems could create genuine security challenges if vulnerabilities are exposed before organizations have time to respond.
Not everyone is fully convinced by the approach, however. Speaking before the order was finalized, policy experts raised concerns about transparency and accountability. Critics argue that while evaluating AI models for cybersecurity threats is a reasonable objective, the process itself remains largely hidden from public scrutiny. Some fear that an unclear review system could eventually give future administrations excessive influence over when and how AI products are released.
Supporters, on the other hand, view the measure as a practical attempt to address emerging threats without imposing sweeping restrictions on innovation. As AI systems become more capable of discovering flaws in software and digital infrastructure, governments and private organizations are increasingly looking for ways to stay ahead of potential risks.
