13 June 2026 will go down in history as the day a Western government treated an artificial intelligence model as a strategic national security asset. Anthropic, the company founded by Dario Amodei and valued at around $965 billion, was forced to completely disconnect its two most advanced models — Fable 5 and Mythos 5 — following a directive from the US Department of Commerce. The order, issued on the afternoon of Friday 12 June, prohibited access to anyone of foreign nationality, including Anthropic’s own employees, which forced the company to withdraw both systems for all its customers.
The trigger was a call from Amazon’s CEO, Andy Jassy, to the White House on Thursday 11 June. According to sources close to the case, Amazon researchers managed to bypass Fable 5’s safeguards and extract information that could be exploited in cyberattacks. In less than 24 hours, the matter escalated to a high-level meeting with Treasury Secretary Scott Bessent, White House Cybersecurity Director Sean Cairncross and Chief of Staff Susie Wiles. By Friday afternoon, Dario Amodei was facing half a dozen senior officials, including Commerce Secretary Howard Lutnick, in a tense conference call. Amodei argued that it was a very specific circumvention method, not a general vulnerability, but the White House was not convinced: the government had been wary of Anthropic for months after the Pentagon added it in March to its supply chain risk list for refusing to hand over its models for mass surveillance and autonomous weapons.
The episode has triggered an earthquake in the artificial intelligence ecosystem. Amazon, which has invested $8 billion in Anthropic and is its main cloud computing partner through AWS, finds itself in a delicate position: it is at once investor, supplier and, now, the actor that provided the security signal that triggered the ban. Meanwhile, OpenAI watches from the sidelines as its main frontier-model rival gets caught in a regulatory maze, and is negotiating with Amazon a possible investment of up to $50 billion. “It is no longer just about who trains the most powerful model”, summed up an industry analyst. “It is about who decides who can use it, and under what conditions”.
— Marta, for intermitente intelligence.
