The U.S. Commerce Department’s national security export controls have expanded beyond limiting model distribution abroad to also restricting access by foreign nationals inside the U.S., according to statements tied to Anthropic’s response. The directive came after research tied to potential safeguard bypass concerns, prompting Anthropic to disable its newest models globally. A cybersecurity CEO, Luta Security’s Katie Moussouris, argued the underlying research was “defense oriented” and not a true “jailbreak,” raising questions about how technical findings translate into broad access rules. Another commentator warned the scope could accelerate “balkanization” of technology access, where students, researchers, and developers face uneven model availability by citizenship. For universities and research centers, these controls have immediate operational implications: procurement and experimentation plans may require nationality- and jurisdiction-aware access controls, and institutional AI risk reviews may need to include export-control compliance—not just data privacy and model safety. The episode underscores how quickly frontier AI governance can shift from policy discussion to concrete restrictions that affect day-to-day research and teaching workflows.