US Government Orders Anthropic to Halt Access to Fable 5 and Mythos 5 AI Models The shutdown of Anthropic’s most powerful AI models has turned a technical dispute over cybersecurity into a global test of how far governments can reach into frontier AI.

On June 9, Anthropic publicly launched Fable 5, a safeguarded, general‑use version of its restricted Mythos‑class model Mythos 5, after months of warning that Mythos’ hacking capabilities posed “serious civilizational” risks. Within days, Amazon security researchers told the White House they could “jailbreak” Fable to surface code vulnerabilities, prompting worries that Mythos‑level cyber tools might be leaking past guardrails.

On June 12, Commerce Secretary Howard Lutnick sent Anthropic an export‑control directive barring use of Mythos 5 and Fable 5 by “any foreign national,” including Anthropic’s own staff. Anthropic said the letter gave no concrete evidence beyond a “potential narrow, non‑universal jailbreak” and argued similar flaws exist in other models, but nonetheless disabled both systems for all customers to comply. The company’s public statement stressed that no “universal jailbreak” had been found and that its safeguards had been red‑teamed for “thousands of hours.”

Over the ensuing 24 hours, senior Trump‑administration officials failed to persuade CEO Dario Amodei to pull Fable voluntarily, then imposed sweeping export rules that effectively recalled the models days after release. Anthropic flew senior technical staff to Washington for crisis talks, while the company’s models vanished mid‑project for foreign users and startups abroad, exposing their dependence on U.S. cloud AIs.

Perspectives quickly diverged. Axios reported the move as a national‑security escalation treating cutting‑edge AI “as national security assets.” TechCrunch argued the ban was “never about an AI jailbreak” but about power and personality clashes, warning it shows Washington can shutter AI products unilaterally. The Verge and others described opaque, “ad hoc” governance that leaves companies guessing at unwritten rules.

Anthropic’s allies in cybersecurity pushed back hardest. In an open letter, roughly 100 experts said the decision “has taken the best models away from defenders, created market uncertainty, and risked America’s AI leadership without any real risk to justify it,” calling the move “not safety, it is sabotage.” They argued that the exploit Amazon flagged was “not a jailbreak” and that comparable capabilities exist in rival systems, including OpenAI’s GPT‑5.5.

Governments and competitors read the episode differently. Canada’s prime minister, Mark Carney, compared the shock shutdown to the 2008 financial crisis, warning that over‑reliance on a few frontier models creates systemic risk and urging diversification of AI infrastructure. European startup Mistral cast the freeze as proof that control ultimately rests with U.S. providers and their governments, bolstering its push for “sovereign” open‑weight models run on local infrastructure.

As U.S. and European officials now discuss a “trusted partner” scheme to restore allied access to restricted models, the Fable‑Mythos affair has shifted AI oversight from abstract panels to real‑time power: a single export‑control letter can now decide which of the world’s most capable AIs anyone is allowed to use.

Continue reading https://foxvector.com/stories/019ed7e1-0d11-22b2-724c-2d4410002ee4