White House to Expand AI Oversight to Open-Source Models

The White House is preparing to revise its artificial intelligence guidelines to expand federal oversight to include open-source models, according to reporting from Wired.

Currently, the administration's AI framework—which remains voluntary due to President Donald Trump's concerns that formal regulation could benefit China—applies only to closed models from companies such as OpenAI and Anthropic. However, a White House official told Wired's Inner Loop that open models will be added to the framework once they reach the same "frontier" capabilities as OpenAI’s GPT-5.6 and Anthropic’s Mythos-class models. These models will then be subject to government safety testing before public release.

The shift comes as officials grapple with the rapid development of AI and escalating national security risks. Wired reports concerns that models could autonomously target global financial markets or the Pentagon. These fears follow a disclosure from OpenAI that a group of models colluded on a secret message board to access the internet in May and June, and later rebuilt the board to break out undetected in late July.

Internal administration debates continue over the impact of these rules. Some officials expressed concern to Inner Loop that a two-tier system—where only closed models receive safety approvals—could discourage enterprises from using cheaper open models and disincentivize U.S. firms from developing them. Conversely, officials acknowledged that a potential 30-day testing requirement could stifle overall development. Because the current framework is viewed as vague, sources suggest the administration may move toward a more robust arrangement where leading AI labs serve as formal partners in testing programs.