← Back to feed

Two weeks after the Commerce Department ordered Anthropic to pull both Mythos 5 and Fable 5 from every user worldwide, the block is partially lifted. Commerce Secretary Howard Lutnick sent a second letter to Anthropic Friday, clearing Mythos 5 for deployment to more than 100 US institutions.

The clearance covers entities named in Annex A of the letter — described as including Fortune 500 companies and government agencies — plus their foreign national employees and Anthropic’s own foreign national employees. Outside that list, the original worldwide export license requirement from June 12 still stands.

Fable 5, the consumer-facing model from the same family, is not mentioned in the new letter. It remains offline under the original order.

“Since the issuance of my June 12 letter, Anthropic has worked with the U.S. government to address risks associated with the Covered Models. These efforts have yielded significant progress,” Lutnick wrote to Anthropic’s chief compute officer Tom Brown. “Anthropic has committed to work with the U.S. government on protocols and standards and releases for its models.”

Anthropic confirmed Friday afternoon: “Since June 12, we’ve been working closely with the US government to restore access to Claude Mythos 5 and Fable 5. Today, the government notified us that Mythos 5, our strongest cybersecurity model, can be redeployed to a set of US organizations that operate and defend critical infrastructure. We’re restoring access for these organizations quickly, and we’re continuing to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.”

What Got Cleared

The new arrangement removes the export license requirement for:

• Companies and institutions on the Annex A list (not yet publicly disclosed)
• Their foreign national employees
• Anthropic’s own foreign national employees

Everything else remains under the June 12 order. Non-listed US companies, all international users, and any foreign nationals not employed by approved entities still require individual export licenses. The government has not signaled it will process those at scale.

The Annex A list skews toward critical infrastructure operators and vetted defenders. Mythos 5 was specifically designed for cybersecurity and biological research applications — its 95.5% SWE-bench Verified score and 88% Terminal-Bench 2.1 result made it the most capable AI coding and security analysis system deployed before the shutdown.

Fable 5’s Absence

Fable 5’s exclusion from the letter is notable. The model was briefly the most powerful AI system in wide consumer release — positioned by Anthropic as the publicly accessible version of the Mythos capability tier before the June 12 order shut it down three days after launch.

Anthropic’s statement commits to seeking Fable 5 restoration but offers no timeline. Infrastructure signals suggest the groundwork is being laid: Amazon Bedrock reinstated Fable 5 in its catalog earlier this week, and the model appeared in Claude Code release notes ahead of any official announcement — moves that typically precede staged rollouts.

The Commerce Department’s silence on Fable 5 leaves Annex A as the only route back. Consumer access and institutional access are now on separate regulatory tracks with different criteria.

The Same-Day Parallel

Friday’s Mythos 5 clearance came the same day OpenAI released GPT-5.6 to a short list of government-approved partners. Two consecutive frontier model launches, two different labs, both gated by a Commerce Department-approved access list. The coincidence of timing suggests a coordinated framework rather than two isolated decisions.

The government-as-gatekeeper model is becoming structural: companies must appear in a federal annex before they can access the most capable AI systems. For international access — which Fable 5’s export controls address most directly — the bar is higher still.

The International Dimension

The two-week shutdown had measurable diplomatic fallout. At the Pax Silica Summit this week, India’s Ministry of Electronics and Information Technology Secretary S. Krishnan raised the abrupt cutoff directly with US officials. US Under Secretary for Economic Growth Jacob Helberg told the summit that for trusted partners “access will not be an issue,” and described a “gradual measured approach” to releasing Anthropic models to allies.

Helberg’s framing points toward a tiered AI export architecture that mirrors Cold War-era technology controls — allies get structured access paths, adversaries do not. India’s intervention indicates the disruption from the June 12 order extended well beyond US borders, affecting foreign national employees at Indian institutions that had integrated Mythos into production workflows.

Three Precedents Set

The June 12 to June 26 episode established a framework that will apply to future frontier model releases:

First, the US government demonstrated it can force an AI lab to disable a product globally, for all users, within days of launch. The mechanism exists and has been used.

Second, reinstatement produces tiered access — not open restoration. The Annex A list determines who gets the model. The list is not public and the criteria for inclusion are not published.

Third, consumer and institutional access are now on separate regulatory tracks. Mythos 5 (cybersecurity use, vetted partners) returned first. Fable 5 (general use, broader access) has no timeline.

For the 100+ institutions now cleared, Mythos 5 restores AI security capabilities that had been operational and then abruptly cut. For Anthropic, the partial restoration secures the revenue from its highest-value enterprise clients while the Fable 5 negotiation continues. For everyone else, the model that scored 95.5% on SWE-bench Verified and 88% on Terminal-Bench remains effectively unavailable.