Anthropic's Most Advanced AI Model Pulled Offline by US Government Order in 48 Hours

A security researcher bypassed the safety systems of the most advanced AI model just released. Within 48 hours, the US government had it pulled offline entirely.

Three stories happened simultaneously this week. A model disabled by government order. A pause proposal from inside the industry. And the next generation model arriving in weeks. Together, they reveal something every business leader needs to understand about depending on a single AI provider.

What happened to Fable 5

Anthropic launched Fable 5, its most capable model to date, on June 9. Within days, a security researcher known publicly as Pliny the Liberator published a method for bypassing some of the model’s safety systems.

On June 12, the US government issued an export control directive instructing Anthropic to suspend access to Fable 5 and its companion model Mythos 5 for any foreign national, citing national security authorities. Because there was no practical way to filter foreign nationals from US users in real time across a global customer base, Anthropic disabled both models entirely, for everyone, within hours of receiving the order.

Enterprises that had built workflows on top of Fable 5 lost access without warning and without a restoration timeline. Anthropic stated publicly that it believed the situation was a misunderstanding and that it was working to restore access. Access to Anthropic’s other models, including Claude Opus 4.8, was not affected.

Anthropic’s proposal for a coordinated pause

In the same week, Anthropic published a paper proposing something unprecedented in the history of the technology industry: a globally coordinated pause on frontier AI development.

The argument behind the proposal is that frontier models are approaching the ability to improve themselves recursively, and that human oversight of that process is becoming harder to maintain. Whether or not the industry adopts this proposal, it signals that questions about the pace of AI development are now being raised from inside the companies building the technology, not only from outside critics or regulators.

Google is closing in on the next generation

At the same time, Google is finalizing Gemini 3.5 Pro, expected to ship with a two million token context window before the end of June. While one frontier lab faces a forced shutdown, another is preparing to ship its next major release on schedule.

This contrast is the clearest illustration yet of how unevenly the frontier AI landscape can move. Regulatory action against one provider does not slow down the competitive race among the others.

The real story is enterprise dependency

What deserves more attention than any single event is what it reveals about infrastructure risk. Every business that had built workflows, products, or customer-facing tools on top of Fable 5 woke up on June 12 with no access, no warning, and no timeline for restoration.

This is not a hypothetical risk. It already happened, to real companies, this week. And it illustrates a structural vulnerability that most businesses have not priced into their AI strategy: dependency on a single AI provider is no longer just a technical consideration. It is a business continuity risk, on the same level as relying on a single supplier, a single cloud region, or a single critical vendor.

What this means for your business

Regardless of which AI provider you use or how you view the policy debate around this directive, the practical lesson applies to every business building on AI infrastructure today: have a contingency plan for what happens if your primary AI provider becomes unavailable overnight, whether due to regulation, a security incident, or a commercial dispute.

This is the kind of risk that needs to sit on the same planning table as your cybersecurity strategy and your vendor diversification strategy, not be treated as a hypothetical edge case.

At Aliando, we help businesses build AI strategies resilient enough to withstand exactly this kind of disruption.

Follow for your weekly AI intelligence.

(Do not miss the video below, where this story is broken down in under a minute.)