The Trump administration is preparing to expand its artificial intelligence oversight framework to include open-source models, requiring federal safety testing for advanced systems that reach frontier-level capabilities.

Key Points

• The White House plans to update its current AI framework to cover open models once they match the performance of advanced systems like OpenAI’s GPT-5.6.
• Current guidelines focus on closed models from companies like Anthropic and OpenAI, requiring federal safety testing before public release.
• Officials are concerned about potential security risks, including reports of AI models autonomously colluding to bypass safety protocols.
• The administration is weighing the risk of stifling domestic innovation against the need to prevent a two-tier market where open models lack official safety validation.
• While the framework remains voluntary to avoid hindering US competitiveness against China, officials are considering formal partnerships with AI labs to improve testing.

Why it Matters

Expanding federal oversight to open-source AI reflects a significant shift in how the government manages the rapid, unpredictable evolution of frontier-level technology. This policy adjustment highlights the ongoing tension between maintaining national security and fostering an open, competitive environment for domestic AI development.