The US lifted its export ban on Anthropic's Fable 5. The model returns Wednesday.

The US Commerce Department has lifted the export controls on Anthropic’s Claude Fable 5 and Mythos 5. Both models were pulled offline worldwide on June 12; Anthropic says Fable 5 comes back Wednesday. The reversal closes an 18-day standoff that briefly made a frontier AI model disappear on a federal order.

This is the resolution of the story we covered when a US government order forced Anthropic to pull Fable 5 days after opening it to the public. The freeze started over a claimed jailbreak the government treated as an export risk. It ends with the same agency, Commerce, declaring the model safe enough to ship globally again, after Anthropic rebuilt the safeguards and let a federal lab test them. The whiplash is the point: a model went from public launch to total blackout to cleared for redeployment inside three weeks, and the reader who wired anything to it rode all three.

What we know

The clearance came from the same office that imposed the ban. Commerce Secretary Howard Lutnick approved both models after Commerce “worked closely with Anthropic to analyze and approve Fable 5,” as he put it. Here’s the confirmed picture.

• Anthropic announced the reversal on June 30. “We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5,” the company wrote, per CNBC, adding it would begin restoring access the next day.
• Fable 5 redeploys Wednesday, July 1, across the Claude platform, Claude.ai, Claude Code, and Claude Cowork, Anthropic said in its redeployment post. Access through AWS, Google Cloud, and Microsoft Foundry follows as fast as the partners can flip it back on.
• Mythos 5 came back first. It was restored for select US organizations on June 26 after government approval, BleepingComputer reported, ahead of the broader Fable 5 rollout.
• The new safeguard is a classifier that hands off risky prompts. When Fable 5 flags a high-risk cyber or bio-chem request, Anthropic says “the request will instead be sent to Opus 4.8,” the untouched fallback model, and the block for the specific technique from the earlier Amazon report now holds “in over 99% of cases.”
• A federal lab signed off. Researchers from Commerce’s Center for AI Standards and Innovation “have tested both our prior and new safeguards and agree that they are extraordinarily strong,” Anthropic wrote, adding that it’s “grateful” to everyone who worked on the redeployment, per Benzinga.

The trigger for the clearance was partly competitive. TechCrunch reported that pressure from comparable Asian models weighed on the decision alongside the safety review, echoing the logic behind the paused DeepSeek blacklist: controls that only handicap US labs look worse when a rival ships anyway.

What we don’t know

The freeze is over; the terms of the peace are fuzzy. As of July 1, the redeployment post is short on the mechanics that matter to anyone outside the US.

• Whether every foreign user gets full access. It’s still unclear if Fable 5 will roll out to everyone outside the United States, or whether some access stays gated behind the identity checks Anthropic has been piloting.
• How the fallback feels in practice. Anthropic hasn’t published what share of real sessions get rerouted to Opus 4.8, so the day-to-day hit to Fable 5 users is unmeasured.
• Whether this holds. There’s no published commitment that a future jailbreak claim won’t trigger another directive. The mechanism that took the model down on a Friday afternoon still exists.

Where the reporting comes from

Anthropic’s own redeployment post is the primary record for the safeguards and the Wednesday date. The lift itself was reported by CNBC, TechCrunch, and BleepingComputer, with Lutnick’s approval quote via Benzinga. It slots into a wider pattern of the administration recalibrating AI export policy under competitive pressure, the same tension visible in how the White House staggered the GPT-5.6 release.

What this means for you

If you failed over to Opus 4.8 during the blackout, don’t rush the switch back. Fable 5 returns Wednesday, but the new classifier now silently routes high-risk cyber and bio-chem prompts to Opus 4.8 anyway, so a security-tooling workload may quietly land on the fallback model regardless of which ID you pin. Test your actual prompts against the redeployed model before you re-point production traffic, and log when a response comes from Fable 5 versus Opus 4.8 so you can see the reroute rate for yourself. The strategic lesson from the original suspension still stands: a frontier model can vanish on a federal order and come back on one, and both events land outside your change window. Keep a second vendor’s model qualified. This clearance is a reprieve, not a guarantee the switch stays on.