White House EO: Early Frontier AI Model Evaluation

The White House has issued an Executive Order directing the U.S. government to establish a classified benchmarking process. This process will evaluate the cybersecurity capabilities of advanced AI systems. AI developers can voluntarily submit their models for evaluation. This could help the government understand emerging cyber capabilities and identify "covered frontier models." The order also suggests a voluntary process for developers to provide federal government access up to 30 days before public release. Federal agencies are also directed to prioritize AI-related cybersecurity initiatives and strengthen national cyber defenses. One expert expressed concern that the approval process could become a way for companies to gain a marketing advantage, potentially shifting focus to lobbying rather than solving security issues. Another expert notes that while early access is sensible, it's voluntary and won't prevent the release of models with advanced cyber offensive capabilities. This initiative highlights the government's concern over the cyber risks posed by advanced AI models.