When Governments Pull the Plug on AI — What the Fable 5 Case Means
Anthropic was forced to shut down its most powerful models after a US government order. A precedent with far-reaching consequences for the entire AI industry.
The World’s Most Powerful AI Model — And Then: Darkness
On June 9, 2026, Anthropic released Claude Fable 5, the first publicly available model in the so-called Mythos class. These models aren’t just a little better — they’re so capable in areas like cybersecurity, biology, and chemistry that Anthropic safeguarded them from the start with separate AI classifiers. Queries on sensitive topics are automatically routed to a weaker model (Opus 4.8).
Sounds reasonable. But just hours after launch, a security researcher claimed to have jailbroken Fable 5 using a multi-agent attack. Anthropic pushed back: some of the demonstrated outputs weren’t even from Fable 5, and the rest contained only publicly available information. The damage, however, was already done.
The Government Steps In
On Friday, June 12, the order arrived: the US government directed Anthropic to suspend access to Fable 5 and Mythos 5 for all foreign nationals, citing national security and export controls. The problem? Anthropic can’t reliably verify the nationality of every user in real time. The only way to guarantee compliance? Shut it down. For everyone. Worldwide.
API calls now return errors, active sessions were terminated, and all services fall back to the weaker Opus 4.8.
Why This Matters for Everyone
This is the first time a leading AI company has been forced to take a publicly deployed model entirely offline due to government intervention. And the consequences extend far beyond Anthropic:
Export control as a regulatory tool. The US government didn’t go through the traditional route of new legislation or agency rulings. Instead, it weaponized export controls — an instrument originally designed for physical goods like weapons and semiconductors. Using it for software distributed globally via APIs is unprecedented.
Dependency becomes risk. Companies worldwide that built their products on Claude APIs suddenly found themselves without their most powerful tool. When a government can pull the plug at any time, “Build vs. Buy” suddenly becomes a question of national sovereignty.
Safety vs. Overreach. Ironically, Anthropic — the company that has championed safety research — got hit by its own messaging. Their own warning blog posts about the dangers of their models were cited by the government as justification for intervention. Safety communication as a double-edged sword.
What We’re Taking Away From This
As a software studio that builds with and on AI tools, we’re watching this unfold with mixed feelings. On one hand, reasonable safety measures are essential. On the other, this case shows how quickly regulatory intervention can become a kill switch for entire technology stacks.
For us, the lesson is clear: diversifying AI providers isn’t a luxury — it’s a necessity. Anyone relying on a single provider carries an incalculable risk — not just from technical outages, but because geopolitical decisions can change everything overnight.
The question that remains: is this the beginning of a new era of AI regulation, or just a one-off? We’ll probably find out sooner than we’d like.