White House Plans to Launch AI Safety Review Framework, Targeting Only Closed-Source Models

According to The New York Times, White House officials held a meeting with top AI companies on Tuesday, announcing that the federal government plans to conduct security risk reviews on specific types of AI models. The review is limited to "closed-source" models that do not disclose underlying code (such as products from Anthropic, OpenAI, and other companies). Open-source models are not currently included in the review, but officials stated that as technology develops, relevant policies may be adjusted.