Trump Signs Executive Order That Invites Vetting of Top AI Models for National Security Risks
President Trump signed an executive order establishing a framework for the federal government to vet the national security risks of advanced AI systems, primarily focusing on models developed by companies like Anthropic and OpenAI. The voluntary process, overseen by the NSA, aims to secure critical infrastructure and strengthen cyber defenses, but faces criticism regarding potential government overreach and the selection criteria for AI models. This follows a previous attempt to implement similar oversight that was scrapped due to concerns about stifling innovation.
The executive order outlines a 30-day review period for AI systems before their public release, intended to mitigate national security risks. Participation is voluntary, with the government focusing on systems developed by companies such as Anthropic (Claude Mythos), OpenAI (ChatGPT), and Google, often referred to as "frontier labs." The White House emphasized that this is not a broad oversight of all new AI models, aiming to avoid "chilling effects" on free speech and innovation. However, concerns remain about the discretion granted to the NSA director in determining which models receive scrutiny, potentially leading to selective targeting of companies in disputes with the administration.
Following Anthropic’s announcement of Claude Mythos and amidst a legal battle with the Pentagon, Treasury Secretary Scott Besant and Jerome Powell convened a meeting with Wall Street CEOs to address potential cybersecurity vulnerabilities. Anthropic has limited access to Claude Mythos, expanding access to 150 organizations. OpenAI welcomed the policy, emphasizing the need for democratic oversight and stakeholder input. Senator Mark Warner criticized the administration's belated recognition of the need for such a directive, highlighting the previous dismantling of similar guardrails.