The Biden administration is preparing a framework for a voluntary review of artificial intelligence models, focusing on those that pose national security risks. This initiative aims to address concerns surrounding the rapid advancement of AI technology and its potential implications.

The review process will specifically target closed-source AI models, meaning those where the underlying code is not publicly accessible. This approach allows for a more controlled assessment of potential vulnerabilities and risks without requiring companies to disclose proprietary information. However, AI models that publish their underlying code will be excluded from this particular review.

This framework represents a significant step by the U.S. government to proactively manage the security challenges presented by sophisticated AI systems. The details of the framework are still being finalized, but the focus on closed-source models suggests an effort to balance innovation with safety.