Behind-the-Scenes Power Struggle Exposed: US Department of Commerce Secretly Obstructed Fable 5 Before Release, Jailbreak Incident Turned into Government's Forced Intervention Opportunity

According to Dongcha Beating's monitoring, the U.S. government's emergency decision to halt Anthropic's Fable 5 and Mythos 5 access permissions was actually the eruption of behind-the-scenes maneuvering that had been going on for weeks. Insider sources revealed that as early as before the official release of the two models on June 9, the U.S. Department of Commerce, led by Secretary Howard Lutnick, had expressed concerns about the models' powerful network security reasoning capabilities and had privately attempted to intervene and obstruct Anthropic from releasing the models, but without success.

To address potential security risks, Anthropic took a compromise approach, releasing to the public only a commercially available version of Fable 5 with extremely stringent filters, while restricting the security-filter-removed "bare model" Mythos 5 under the tight control of "Project Glasswing," opening it only to a very small number of red team organizations and defense departments that passed background checks.

However, just three days after the models were released, a third-party entity claimed to have successfully "jailbroken" the controlled Mythos 5 and reported this. This immediately became a perfect opportunity for Lutnick and the Department of Commerce to forcefully intervene. The Department of Commerce not only sent a formal letter to Anthropic overnight but also, through export control means, unprecedentedly directly froze the commercial access rights of specific models.

The industry generally points out that the Department of Commerce this time wielded the export control stick directly at the model itself, not only setting an extremely dangerous regulatory precedent but also indicating that the U.S. government is shifting towards a unilateral and interventionist approach to AI security regulation.