Can Claude Fable 5 be unblocked? Presidential Technology Advisor: Fixing one issue can remove the restrictions, letting the “fable” go back online

The U.S. government issued an export control order against Anthropic yesterday (13th), citing national security reasons, prohibiting any foreign nationals from accessing the latest models Fable 5 and Mythos 5. Presidential Technology Advisor David Sacks explained the reason for this ban.
(Background summary: a16z founder Marc Andreessen’s view on U.S. AI regulation — opposing laypeople setting arbitrary rules and welcoming beneficial "brakes")
(Additional context: Anthropic CEO has only one direct subordinate, overturning industry norms)

Table of Contents

Toggle

• Sacks’ reasoning behind the accusations
• Anthropic’s rebuttal: Is a narrow jailbreak worth recalling the model?
• Who holds the power over deployment and life-or-death decisions

Anthropic, proud of its "AI safety" credentials, saw its latest models Fable 5 and Mythos 5 completely cut off from all users worldwide over a single export control order issued over the weekend. According to official Anthropic announcements, the U.S. government invoked national security authority, requiring the company to prohibit any foreign nationals—regardless of whether they are inside or outside the U.S., including Anthropic’s own foreign employees—from accessing these two models, effective immediately and without a grace period.

Because technically it’s nearly impossible to "block only foreigners," Anthropic stated that the only compliant solution is to shut down access for all customers globally (including U.S. citizens), with the rest of their models unaffected. This is also the first time the U.S. government has used export controls on a mainstream AI model that is already commercially deployed and serving hundreds of millions of users, citing national security concerns.

Sacks’ reasoning behind the accusations

On the morning of the 14th, David Sacks, co-chair of the U.S. President’s Council of Advisors on Science and Technology (PCAST), publicly recounted the conflict on X. Sacks said, "Fable is basically Mythos with safety rails," and both are essentially the same model.

According to Sacks, a highly trusted partner, trusted by both Anthropic and the U.S. government, found a way to bypass the safety rails (meaning inducing the model to circumvent built-in restrictions) during testing of Fable.

The government demanded that Anthropic CEO Dario Amodei choose between "fixing the jailbreak" and "removing the model from service." Dario refused both options, opting to keep Fable online. Sacks questioned Anthropic’s blog claim that the jailbreak was "not serious," asking: how could a jailbreak that enables cyber weapons to operate not be serious? He believes this contradicts Anthropic’s long-standing claim of being an "AI safety company."

He sarcastically noted that Anthropic previously claimed Mythos was a "cyber weapon" that should be regulated as a weapon.

I’ve had a number of conversations with folks inside and outside government about the current situation with Anthropic, and here is what I believe to be true:

— As we know, Anthropic publicly released its Mythos class models earlier this week under the commercial name Fable.…

— David Sacks (@DavidSacks) June 13, 2026

However, Sacks emphasized that the government issued the order "reluctantly," and hopes that Anthropic will quickly fix the issues, lift the restrictions, and fully relaunch Fable, implying "the ball is in Anthropic’s court."

Anthropic’s rebuttal: Is a narrow jailbreak worth recalling the model?

Anthropic also responded in an official statement, pointing out that the jailbreak cited by the government is "narrow and non-general," essentially just asking the model to read a specific piece of code and find software flaws, a capability that cybersecurity defenders use daily.

In simple terms, it’s not a universal key that can open all defenses, but a spare key that opens only one door.

Anthropic emphasized that Fable 5 already prioritizes cybersecurity as a key defense layer, employing a "defense-in-depth" strategy: besides the model’s own jailbreak resistance, they also conduct red-team testing in collaboration with the U.S. government, and retain customer data for up to 30 days for monitoring. The official further pointed out that the same vulnerabilities exist in other publicly available models on the market (including OpenAI’s GPT-5.5), but those models have not been subjected to the same national security export controls.

This comparison serves as a public question: If standards are consistent, why is only Anthropic being regulated?

Anthropic concluded that it will comply with the legal order but does not agree with it, calling for any takedown decisions to follow a "transparent, fair, and technically justified legal process," rather than a letter that causes all global customers to go offline collectively at dusk.

Who holds the power over deployment and life-or-death decisions

The deeper core of this dispute is the issue of "who defines the severity" — the authority to decide. Anthropic says: technically, it’s just a narrow vulnerability, insufficient to recall a commercial model serving hundreds of millions. The U.S. government says: as long as the safety rails can fail, the weapon is exposed.

Behind these two risk assessment frameworks are two fundamentally different worldviews: one sees Fable as a flawed but statistically safe product, the other as a weapon requiring zero-tolerance protection.

When a national government can use "national security" as a one-click switch to shut down its leading AI flagship model, who truly controls the deployment of closed-source models?