Industry Revolt Over Fable 5 Ban as Amodei Talks Stall

What happens when a company’s two strongest AI models vanish from the market overnight? That is exactly what Anthropic is facing now. On June 12, the US Department of Commerce issued an order barring the Fable 5 and Mythos 5 models from being made available to foreign nationals, both inside the US and abroad. Six days later the models are still offline, with no announced timeline for their return.

The industry pushed back fast. More than 100 cybersecurity leaders signed an open letter calling for the decision to be reversed. Their argument is simple: in their view the ban does more harm than good, because it cuts off legitimate users from tools they already rely on in their daily work.

How did the ban happen in the first place?

The trigger was an alleged jailbreak method that Amazon chief Andy Jassy escalated to Treasury Secretary Scott Bessent. Bessent passed the finding along, and Commerce Secretary Howard Lutnick then issued an export-control directive. In a few short steps, a technical issue raised by one company became an official government measure.

Anthropic disputes the risk assessment. It argues that the method only exposes minor, previously known vulnerabilities that are also accessible through OpenAI’s rival GPT-5.5 model. And that model, to make matters more interesting, is not under any ban. If the danger is real, Anthropic asks, why does the measure apply only to its own models?

CEO Dario Amodei is trying to fix the situation directly. He held several calls with administration officials, and a senior team traveled to Washington. There is still no agreement, however. By all accounts, tensions only eased somewhat on the sidelines of the G7 summit, but a concrete date for the models’ return still does not exist.

What does this mean for users and the market?

For those of us who use these tools in our daily work, the message is uncomfortable. Access to a serious AI model can disappear not because of a failure or a business decision, but because of a government move made in a few steps, from a single email to an official directive. That is a level of uncertainty no team wants in its workflow.

Anthropic, meanwhile, is working on two fronts. On one, it is pushing legally and diplomatically to get the ban lifted. On the other, it is expanding into new markets, having opened an office in Seoul and saying it expects the models back within days. Whether that is a realistic estimate or an attempt to calm user concerns will become clear soon.

For now the main question remains, one that troubles both the industry and ordinary users: if a single jailbreak can pull two leading models off the market indefinitely, how stable is the foundation we are building on commercial AI tools?