Anthropic's Fable 5 Ban: US Government's Unprecedented AI Restriction Triggers Industry Turmoil, Legal Battles

The US government’s directive to Anthropic, issued on June 12th, has effectively disabled access to its highly capable Fable 5 and Mythos 5 AI models for all foreign nationals, regardless of their location or affiliation, including Anthropic’s own international employees. This abrupt action, coming just three days after the models’ release, was reportedly prompted by concerns over alleged reselling of access to sanctioned regions (specifically China via SK Telecom, a South Korean investor), and reports of a “jailbreak” capability. Anthropic, while complying, argues that the alleged “jailbreak” – the models’ ability to analyze and fix codebase flaws – is an intended and essential feature for developer tools, asserting that similar capabilities exist in other frontier models like OpenAI’s GPT 5.5. Despite initial promises from Anthropic regarding swift resolution and transparent communication from the government, 11 days have passed without the models’ reinstatement or a clear timeline for their return, with negotiations reportedly stalled.

This unprecedented restriction has sent ripples through the AI industry, eliciting strong reactions from various stakeholders. A US-based litigation tech company, Legion, a customer heavily reliant on Fable 5, has filed a 43-page lawsuit against the US government and its agencies. The lawsuit challenges the legality of the directive under the International Emergency Economic Powers Act, arguing it unlawfully prohibits the export of informational materials. Concurrently, a bipartisan group of US Congress members has formally requested greater transparency from the Department of Commerce, expressing concern that this action sets a dangerous precedent for frontier AI regulation. Industry experts worry that by restricting advanced proprietary models while open-weight alternatives like GLM 5.2 rapidly approach similar capabilities, the government inadvertently incentivizes the use of potentially less regulated models, raising critical questions about US leadership in AI and the future of innovation in the sector.