White House AI Review Targets Closed Models Only
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- White House officials told top AI companies in a Tuesday meeting the federal government plans to review only "closed" AI models for security risks — those that don't publish their underlying code.
- Closed-source models from companies like Anthropic and OpenAI fall within the review's scope, while models that publish computer code for public download and modification will not be reviewed, though that could change as the technology advances.
- The framework codifies U.S. government oversight of leading AI labs, giving Washington greater ability to weigh in on and potentially slow the rollout of the technology.
- The policy defers addressing "increasingly powerful open-source models made by Chinese companies" that pose potential security threats, according to people familiar with the discussions.
- Excluding open-source models from review could benefit companies like Meta that use publicly available code to build products competing directly with Anthropic and OpenAI.
- The announcement follows months of private discussions between Trump administration officials and a handful of the world's leading tech companies.
Why it matters: The framework gives the U.S. government explicit, formalized oversight over the dominant closed AI labs — Anthropic and OpenAI — while leaving unregulated the open-source models, including those from Chinese firms, that could undercut them. By scoping review to closed models only, the administration effectively grants catching-up competitors like Meta, which build on open-source, a regulatory head start.
