The White House is preparing an executive order that would ask advanced AI labs to provide their models to the government for a 90-day review before public release.
The White House is preparing an executive order that would ask advanced AI labs to provide their models to the government for a 90-day review before public release.

U.S. President Donald Trump is expected to sign an executive order as soon as Thursday creating a voluntary framework for federal oversight of advanced artificial intelligence models, requiring a 90-day pre-release review period for systems deemed "covered frontier models."
According to a draft of the executive order circulated by the White House, the proposal would ask AI laboratories to hand over their models to the federal government for cybersecurity and capability reviews three months before any public release. While the framework is voluntary, it signals a more assertive federal posture on AI governance.
The order contains two main components: one focused on strengthening cybersecurity protections for Pentagon systems by encouraging information sharing with AI companies, and a second that establishes a process for defining and reviewing "covered frontier models." The 90-day timeline is a significant change for an industry accustomed to rapid development cycles.
The executive order could introduce new compliance burdens and development delays for major AI labs like OpenAI, Google DeepMind, and Anthropic. For investors, the move introduces a new regulatory variable, with the definition of "covered frontier models" being a critical factor in determining the framework's ultimate impact on the sector's competitive dynamics.
The first part of the draft order is aimed at protecting Pentagon systems. It encourages AI companies to share information with federal agencies, effectively asking them to disclose any vulnerabilities their models might discover in U.S. defense infrastructure. The concern is that advanced AI could be a dual-use tool, capable of identifying and exploiting weaknesses in critical systems. The Pentagon's interest is in understanding these capabilities before they become widely available, preventing adversaries from using public model releases to generate novel cyberattacks.
A major consequence of the order will be the process of defining which AI systems qualify as "covered frontier models." This definition will determine which models are subject to the pre-release review. If the definition is too broad, it could stifle innovation by creating bureaucratic hurdles for models that pose little risk. If it is too narrow, the most powerful and potentially dangerous systems could evade scrutiny. The draft order proposes creating a mechanism to establish this definition, a move that will be closely watched by the industry. While the framework is voluntary, companies that choose not to participate may face reputational damage and could invite more stringent, mandatory regulations in the future. The 90-day review period, while a significant ask, provides the administration with a tool to apply soft pressure on labs to comply. This approach echoes regulatory strategies in other sectors, where voluntary frameworks often pave the way for future legislation.
This article is for informational purposes only and does not constitute investment advice.