The United States government has introduced a new policy aimed at tightening coordination with leading artificial intelligence developers before their most powerful systems reach the public.
President Donald Trump has signed an executive order establishing a voluntary framework that allows major AI companies to share their most advanced models with the federal government up to 30 days before public release.
This means companies such as OpenAI, Google, and Anthropic may provide early access to cutting-edge AI systems so that government agencies can evaluate potential risks, particularly in areas like cybersecurity and national infrastructure protection.
The move comes amid growing concern over highly advanced AI models capable of identifying weaknesses in sensitive systems, including those used by banks, hospitals, and government networks. One of the key catalysts behind the order was concern surrounding Anthropic’s unreleased “Mythos” model, which reportedly demonstrated strong capabilities in exposing system vulnerabilities.
The final 30-day review window represents a compromise after internal debate. Earlier drafts reportedly suggested a much longer 90-day review period, while technology companies pushed for a shorter 14-day timeline, arguing that excessive delays could slow innovation and weaken global competitiveness.
Industry leaders have responded with cautious support. OpenAI CEO Sam Altman said the framework “gets the balance right,” emphasizing the need for both innovation and safety while ensuring trusted defenders can access powerful tools. Google’s Kent Walker also described the order as an important step toward strengthening national cybersecurity readiness. Anthropic similarly welcomed the policy, calling it a positive development for US leadership in AI.
However, the decision has also sparked debate within policy and tech circles. Critics argue that voluntary frameworks may not be strong enough to address the risks posed by rapidly advancing AI systems. Some experts believe that governments should have stronger authority to intervene when models present serious national security concerns.
The order follows a broader shift in US AI policy direction. It replaces earlier approaches that focused on lighter oversight and voluntary compliance, including measures introduced during the previous administration. It also contrasts sharply with the European Union’s more regulatory approach, where binding rules require strict transparency and safety obligations for high-risk AI systems.
Inside Washington, the policy process has not been without tension. Reports suggest internal disagreements and lobbying from tech leaders influenced the final structure of the order, reflecting ongoing pressure between innovation speed and national security safeguards.
Supporters of the policy believe it strikes a practical balance. Critics remain cautious, warning that voluntary cooperation alone may not be enough in a rapidly evolving technological landscape where AI capabilities are advancing faster than regulation.
In the end, the decision reflects a larger global reality: nations are racing not just to build the most powerful AI systems, but to decide who gets to see them first, and who gets to control the risks they bring.
Sometimes, the future of technology isn’t only about how fast it moves—but about who gets to look under the hood before it does.
