Key Takeaways

  • President Donald Trump approved a scaled-back AI order that shortens pre-release model reviews to 30 days.
  • The move aligns with his administration's preference for lighter federal intervention and greater industry self-governance.
  • Growing enterprise reliance on generative AI and emerging standards like the NIST AI RMF shape how businesses respond to the policy shift.

President Donald Trump signed a revised artificial intelligence executive order on Tuesday that trims back the level of federal scrutiny originally under consideration. The decision closes a turbulent chapter inside the administration, one defined by internal disagreements over how tightly the federal government should police rapidly advancing AI systems. The White House moved ahead quietly, signing the order in private rather than staging the ceremony that had been planned for May 21 before Trump rejected an earlier draft.

The final version centers on voluntary participation. Companies releasing powerful new AI models will be asked to submit them for a government review 30 days before launch. That is a noticeable reduction from the 90-day lead time the White House previously explored. Some industry voices had argued for an even shorter 14-day period, seeing longer windows as a drag on innovation.

The shift reflects broader themes in Trump’s AI policy. Analysts at the Brennan Center have noted that Trump’s overall approach narrows the scope of federal oversight compared with the Biden-era EO 14110, placing more weight on deregulation, procurement preferences, and limiting state-level rules. Their assessment, cited in the Brennan Center’s analysis, is that these moves lean on signaling and funding rather than strict enforcement. That said, the federal procurement environment may still nudge commercial vendors toward meeting expectations around "unbiased" or "non-ideological" AI, especially in government-facing markets.

The latest order also lands during a moment of intense AI adoption across industries. According to Gartner, more than 80% of enterprises are expected to use generative AI APIs or deploy generative AI applications in production by 2026, a surge from less than 5% in 2023. That level of uptake means policy shifts like this one ripple through product development timelines, risk functions, and vendor selection strategies.

As enterprises accelerate AI integration, the frameworks guiding safe and reliable deployment matter even when regulations soften. NIST’s AI Risk Management Framework, or AI RMF 1.0, has gained traction across sectors, with many organizations adopting it as a reference point for internal governance. NIST reports strong industry alignment activities, which is one reason major vendors like Microsoft, Google, and OpenAI increasingly anchor compliance around that standard and the related ISO/IEC 42001 management system. When the federal government offers a lighter regulatory touch, companies often lean harder on these voluntary norms, partly for market trust and partly to manage cross-jurisdictional differences.

Those differences are becoming more important. Trump’s stance contrasts with the more prescriptive EU AI Act, and it reshapes the balance of responsibility between federal agencies, states, and industry players. The Brennan Center warns that Trump’s federal preemption tools remain limited, meaning state-level AI laws are unlikely to disappear. For B2B teams operating in regulated sectors, this adds nuance: lighter federal mandates do not erase the patchwork of state rules or international compliance expectations.

Inside the White House, conflicting views on this issue have been evident for months. Some advisers push for minimal regulation to maintain an advantage over China in AI development. Others emphasize the cybersecurity risks introduced by increasingly capable models, such as Anthropic’s Mythos, which officials believe can probe even highly secured systems. Before Tuesday’s signing, the administration convened a small, high-level meeting on Monday to finalize next steps. That structure, smaller and more private than previous gatherings, suggests a desire to avoid the eleventh-hour reversal that occurred on May 21.

That earlier derailment remains a point of tension. According to officials familiar with the process, Trump had been prepared to sign the 90-day version of the order until a call from former AI czar David Sacks prompted a reconsideration. At that moment, the order had already been reviewed by OpenAI, Anthropic, and Google, and had received internal signoff from senior White House officials. This back-and-forth underscores how unsettled the administration’s AI posture has been, even as the technology’s enterprise footprint expands.

IDC estimates global spending on AI systems will reach around $300 billion by 2026, with growth above 25% over five years. That trajectory hints at why voluntary review windows still matter. If developers expect scrutiny but not prescriptions, the incentives often tilt toward managing reputational and security risks early in development. Risk teams at enterprises frequently assess whether their suppliers follow frameworks like the NIST AI RMF or ISO/IEC 42001, since those markers offer a common language for due diligence during rapid adoption cycles.

One lingering question is whether a voluntary 30-day review will meaningfully influence cybersecurity outcomes. Voluntary programs can encourage participation when relationships between government and industry are strong, though participation varies. For companies already coordinating closely with federal agencies, the process may fold into existing internal checks. Others may see the review as a signaling mechanism rather than a substantive requirement. And with model release cycles accelerating, some observers wonder if even 30 days will remain practical in a market where models evolve weekly.

The order sends a clear message about where Trump wants AI oversight to go. It fits a pattern of restricted mandates, greater reliance on industry judgment, and an ongoing strategy to reduce what the administration sees as friction points in the competition with China. Whether that approach holds up as models like Mythos grow more capable and widely deployed is a dynamic the industry will continue to monitor.