On June 13, 2026, the US government issued an export control directive that, for the first time ever, specifically targeted a large language model – and within hours, Anthropic's Fable 5 and Mythos 5 vanished from the internet, even for the company's own employees in the UK, Canada, Australia, and New Zealand.
No Warning, No Exceptions – Fable 5 Goes Dark
The directive bans any foreign national from accessing those models, period. That includes Anthropic employees who aren't US citizens. Fable 5 had been live for just three days; Mythos 5 was still in limited partner preview. Anthropic complied immediately, pulling both models globally. Isaacus, an Australian legal AI firm with clients across multiple continents, called the move "the first time the United States has issued an export control directive for LLM access, ever."
The ban doesn't just affect users. It targets the personal nationality of anyone touching the model, regardless of physical location. That means the US is now asserting that a software model running on US servers can be denied to its own developer's staff based solely on their passport. If you depend on a US-hosted LLM for anything critical, your access can be revoked on zero notice.
Why Sovereign AI Isn't Just a Buzzword
Isaacus has been shipping fully self-hostable, air-gapped models from day one. Not because they're altruistic – they admit it's commercially motivated. As a two-person founding team, offering self-hosting landed them enterprise deployments with multiple Australian government departments. "We simply enjoy disrupting monopolies," they write.
The broader point: any mission-critical AI dependency on a third-party API introduces a single point of political failure. Isaacus points out that even acquisitions can kill access – they note Anthropic acquired Stainless, an SDK generation platform used by OpenAI, Google, and Isaacus, and then shut it down immediately. Export controls are just the most dramatic version of that risk.
What This Enables Next
Isaacus is doubling down. They plan to make not just their models but their entire platform – including the Blackstone Graph and Isaacus Research – fully self-hostable. If the US can switch off Anthropic's flagship models with a single directive, sovereign AI isn't a philosophical debate anymore; it's the only way to keep mission-critical legal intelligence running when the next export order lands.
Source: Our response to the US ban on Fable 5 and Mythos 5
Domain: isaacus.com
Comments load interactively on the live page.