U.S. requires OpenAI phased releases, 2nd major AI firm under review

The White House has ordered OpenAI to stage the release of its most advanced AI models, requiring government safety reviews before wider deployment.

The White House ordered OpenAI to stage releases of its most advanced AI models, requiring government safety reviews before wider deployment — a regulatory clampdown that could reshape how frontier AI reaches the market.

"This phased approach allows independent evaluators to identify dangerous capabilities before they reach millions of users," a White House official familiar with the matter said, speaking on condition of anonymity because the discussions are private.

The requirement applies to OpenAI's upcoming frontier models, including successors to GPT-5. Under the framework, each model version must pass safety benchmarks set by the U.S. Artificial Intelligence Safety Institute before receiving clearance for broader release. The timeline for each review phase has not been disclosed.

For OpenAI, the phased release mandate could extend the gap between internal capability milestones and commercial deployment by weeks or months, potentially affecting Microsoft's integration of OpenAI models into Azure and Copilot products. The move follows similar White House requests to Meta for pre-release evaluation of its Llama models, showing a broader regulatory push across the industry.

Regulatory Precedent Takes Shape

The OpenAI directive follows the White House's formal request to Meta earlier this year to submit its Llama 3 and subsequent models for independent government review before wider release, according to TechRadar. Together, the two actions show the Biden administration moving beyond voluntary industry commitments toward structured oversight of the most capable AI systems.

The reviews are expected to examine dual-use capabilities — whether a model can assist in developing biological weapons, executing cyber attacks on critical infrastructure, or generating sophisticated disinformation at scale. Evaluators also measure tendencies toward deception and other concerning behavioral patterns that emerge during extended interactions.

What Phased Release Means for the AI Economy

For investors, the regulatory shift introduces a new variable into AI valuation models. OpenAI, valued at more than $150 billion in its most recent funding round, now faces potential delays in monetizing its most advanced capabilities. Microsoft, which has invested more than $13 billion in OpenAI and integrated its models across Azure, Office, and GitHub Copilot, may see a slower pace of product upgrades tied to new model releases.

The phased approach also creates a template that other jurisdictions could adopt. The European Union's AI Act already classifies systems by risk level, while China's regulatory framework focuses on content control. The U.S. approach — combining pre-release review with phased deployment — could become a middle path that other nations follow.

For the broader AI sector, the question is whether regulatory oversight will slow the pace of capability advancement or simply shift where and how models are tested. Companies with established safety evaluation pipelines and government relationships may face less disruption than those racing to deploy without structured review processes. The coming months will determine whether phased release becomes the industry standard or a point of competitive friction between U.S. and non-U.S. AI developers.

This article is for informational purposes only and does not constitute investment advice.