THE LONG VIEW

Bottom Line Up Front: The U.S. faces a critical governance gap in frontier AI oversight, as rapid model advancements outpace legal and institutional capacity, risking reactive rather than proactive regulation. Threat Identification: The core threat is the absence of a scalable, legally grounded framework for evaluating high-risk frontier AI models before public release. Recent reports indicate the Trump administration is moving toward mandatory vetting, but the executive branch currently lacks clear statutory authority to enforce such requirements [Lawfare, 2026]. Probability Assessment: High likelihood of attempted pre-deployment controls by 2027, driven by high-profile models like Anthropic’s Mythos reshaping political risk perception. However, legal challenges to executive overreach are probable, making a durable, congressionally authorized regime necessary but not imminent (estimated 60% probability by 2028). Impact Analysis: Without effective oversight, unsafe or strategically misaligned models could be deployed at scale, enabling misuse in disinformation, cyber operations, or autonomous decision-making. Conversely, poorly designed regulation risks stifling innovation or centralizing control without accountability. The stakes include national security, economic competitiveness, and public trust [Rozenshtein & Frazier, Lawfare 2026]. Recommended Actions: 1) Congress should establish clear legal authority for risk-tiered AI oversight; 2) Expand and pilot voluntary evaluation frameworks (e.g., CAISI and the 'kick the tires' model) to build technical capacity; 3) Create a cross-agency AI Resilience Fund to support rapid response to emerging threats; 4) Develop international norms to prevent uncoordinated or oppressive regulatory fragmentation. Confidence Matrix: Governance gap – High confidence; Legal authority gap – High confidence; 2027 pre-deployment push – Medium-High confidence; Effectiveness of voluntary frameworks – Medium confidence (based on current pilot designs).