In a meeting on Tuesday with top artificial intelligence companies, White House officials said the federal government plans to review only certain types of AI models for potential security risks - specifically "closed" models that keep their underlying code private. The decision, which exempts open-source models for now, gives the Trump administration a more direct role in overseeing leading AI labs like Anthropic and OpenAI, but leaves open questions about how to handle rapidly evolving open-source technologies from Chinese firms.
Closed models get federal scrutiny, open source does not
The voluntary review process will cover closed-source AI models, like those built by Anthropic and OpenAI, according to three people familiar with the discussions who spoke on condition of anonymity. The administration will not review open-source models - whose code is publicly available to download and modify - though officials said that could change as the technology advances.
The framework comes after months of private talks between White House officials and major tech companies. It formalizes government oversight of leading AI labs, giving the U.S. the ability to weigh in on deployment timelines. Companies like Meta, which use open-source software to compete with Anthropic and OpenAI, stand to benefit from the exclusion of open models.
Regulatory gap on Chinese open-source AI
The decision delays addressing what some experts see as a pressing security concern: increasingly powerful open-source models from China. By excluding open-source systems from the review, the framework sidesteps one of the most debated regulatory issues in AI today. Administration officials indicated the policy could evolve, but gave no timeline for expanding the review.
The framework is likely to shape the global AI race. Until now, companies largely decided on their own whether to release advanced models. The new U.S. process gives the government a greater opportunity to slow or condition the rollout of closed-source AI, according to the people familiar with the meeting.
Why this matters for government professionals
For officials working on AI policy, national security, or procurement, this framework signals that the federal government will focus its limited review resources on proprietary models from leading U.S. labs. That means government agencies evaluating AI tools should understand whether a given system is closed or open source, since only closed models will face federal security screening. Professionals involved in AI governance may benefit from deepening their understanding of the differences between these model types. An AI Learning Path for Policy Makers and resources on AI for Government are available to help build that expertise.
Your membership also unlocks:
