The US Government Banned Claude Fable 5. Today Anthropic Brought It Back.

Horia Stan is a music producer and sound engineer at The One Records in Bucharest who has used Fable 5 daily since its June 9 launch, including through the three weeks it was pulled from every account worldwide. Today, July 1, 2026, Anthropic restored global access to Claude Fable 5 and Claude Mythos 5 across Claude.ai, the Claude Platform, Claude Code, and Claude Cowork. Most people never noticed the ban happened, or why. Here is the timeline, with sources.

Wait, Fable 5 was banned?

Yes. On the evening of June 12, 2026, Anthropic disabled Claude Fable 5 and Claude Mythos 5 for every customer on earth (Anthropic, 2026). The company received a US government export-control directive at 5:21 PM ET that day ordering it to cut off both models for any foreign national, anywhere, including Anthropic's own non-citizen staff.

Anthropic has no reliable way to verify a user's nationality in real time. Rather than risk non-compliance, it shut both models down for everyone, US citizens included, three days after launch.

What actually triggered the ban

Amazon security researchers had found a jailbreak in Fable 5: a prompt sequence that got the model to bypass its safeguards and identify a number of software vulnerabilities (Tom's Hardware, 2026). In one case, Fable 5 went further and wrote code demonstrating how the flaw could be exploited. That report reached the US government, which classified the capability as a national-security export-control matter.

The detail that undercuts the panic: Anthropic's own testing found that Opus 4.8, GPT-5.5, and Kimi K2.7 could all identify the same vulnerabilities Fable 5 found, and every model tested could produce the same exploit demonstration (The Hacker News, 2026). Fable 5 was not uniquely dangerous. It was just the newest and most scrutinized model when the report landed.

The fix: a single classifier, trained in under three weeks

Anthropic's response was narrow, not a capability rollback. The company trained a new safety classifier tuned specifically to detect and block the exact jailbreak technique from the Amazon report. Anthropic says the classifier stops that technique in more than 99% of attempts as of its June 30 write-up (Anthropic, 2026).

Alongside the classifier, Anthropic opened a HackerOne bug-bounty program specifically for new Fable 5 jailbreaks and committed to giving the US government early access to test future frontier models before public release. The Trump administration lifted the export restrictions on June 30 (Axios, 2026), clearing the way for today's global redeployment.

What comes back today, and on what terms

Fable 5 and Mythos 5 are both back as of July 1, globally, on Claude.ai, the Claude Platform, Claude Code, and Claude Cowork. For Pro, Max, Team, and select Enterprise plans, Fable 5 usage counts against up to 50% of weekly limits through July 7. After that window closes, further use draws from paid usage credits like any other model tier.

I kept my Fable 5 threads open through the entire three-week gap, and the model simply routed silently back to Opus 4.8 mid-conversation on June 12 with no warning in the UI. If you were mid-project when the ban hit, expect to re-verify anything Fable 5 touched right before the cutoff - not because the output was wrong, but because you cannot tell in hindsight which model actually generated it.

Does this change whether Fable 5 is worth using

Not on capability. The SWE-Bench Pro lead over GPT-5.5 that made Fable 5 worth switching to in June has not been revised, and the new classifier targets one specific jailbreak pattern, not general functionality. For coding-heavy workflows, today's return is good news with no real asterisk.

On trust, it is a fair question to sit with. A frontier model got pulled by a government order within 72 hours of shipping, over a jailbreak that every competitor model was equally capable of executing. That says less about Fable 5 specifically and more about how fast export-control law is having to move to keep up with model releases - a dynamic that is not going away with the next release either.

Frequently Asked Questions

Is Claude Fable 5 back today, July 1, 2026?

Yes. Anthropic restored global access to Claude Fable 5 and Claude Mythos 5 on July 1, 2026 across Claude.ai, the Claude Platform, Claude Code, and Claude Cowork, after the US government lifted the export-control order from June 12 (Anthropic, 2026).

Why was Claude Fable 5 banned in the first place?

A US government export-control directive issued June 12, 2026 ordered Anthropic to restrict Fable 5 and Mythos 5 access for all foreign nationals worldwide. Since nationality could not be verified in real time, Anthropic disabled both models for every user globally rather than risk non-compliance.

Was Fable 5 more dangerous than other AI models?

Not according to Anthropic's own testing. Opus 4.8, GPT-5.5, and Kimi K2.7 could all identify the same software vulnerabilities that triggered the export-control order, and every tested model could produce the same exploit demonstration Fable 5 did (The Hacker News, 2026).

Is Fable 5 free after the July 1 relaunch?

Partially, and only briefly. Pro, Max, Team, and select Enterprise plans get Fable 5 counted against up to 50% of weekly usage limits through July 7, 2026. After that date, further use draws from standard paid usage credits.

Did Anthropic change how Fable 5 works?

Anthropic added one new safety classifier that blocks the specific jailbreak technique identified by Amazon researchers, stopping it in over 99% of attempts. The model's underlying capabilities and benchmark performance were not otherwise changed.