The U.S. government rolled back recent export controls that had required Anthropic to disable access to its Mythos and Fable frontier AI models. The reversal was staged—first for Mythos, then later for Fable—after the controls had constrained users globally. The development was framed as a relief for Anthropic and investors, and for users seeking to run autonomous tasks with Fable. Coverage also pointed to cybersecurity implications, noting defenders wanted access to models to find and patch security vulnerabilities before attackers could use comparable capabilities. The episode, however, underscores persistent policy uncertainty: the U.S. continues to operate what observers characterize as an ad hoc licensing regime for frontier models, while formally denying it is a licensing system. A separate report suggests the U.S. may shift toward “voluntary standards” for frontier labs, particularly on cybersecurity. For universities using frontier AI systems and for research labs collaborating with industry, the key issue is compliance volatility—how quickly policy constraints can change, and how that affects model accessibility, experimentation, and risk management.