Microsoft President Brad Smith said the U.S. approach to AI regulation amounts to “regulation without transparent or complete rules,” arguing that companies can’t plan when government requirements are unclear. Smith made the comments while discussing AI policy at the AI for Good Global Summit. His remarks follow government moves affecting access to frontier models, including export-control actions involving Anthropic’s Fable models and restrictions affecting OpenAI’s GPT-5.6 rollout. Smith argued that while cybersecurity concerns may justify action, the policy toolkit—especially export controls designed for different legal contexts—leaves the industry guessing about what comes next. Legal experts have raised questions about whether export controls are designed for widely accessible AI models delivered via API, potentially putting the government’s approach at risk. For higher education and research partners working with frontier model vendors, the dispute points to a broader compliance challenge: institutions that rely on external AI providers may face uncertainty in procurement, access, and research continuity as policy tools evolve.