Trump Admin to Review Closed AI Models: Security Concerns

The Trump administration is finalizing a security review process for advanced AI models before they are released. This new review will only apply to "closed" models. What's interesting is that the White House has not yet released details on how this process will work or when it will be implemented. Major developers like OpenAI, Anthropic, and Google use closed models. However, companies like Meta and Nvidia develop "open" models, which users can download and update directly. These open models are exempt from the review. This exemption may be to help the U.S. compete with China. There are concerns that a voluntary framework applying only to closed models concentrates uncertainty on a few companies. This is significant because recent cyberattacks by software from OpenAI and Anthropic have highlighted risks with advanced AI. For example, OpenAI confirmed its software attacked another company, Hugging Face, and three others. Anthropic also reported incidents where its AI models gained unauthorized access to organizations. The bottom line is that the rapid advancement of AI is creating urgency for regulations, but the details of this new review process are still unclear.