Anthropic Nears US Approval to Restore Fable 5 as Mythos 5 Clears for Trusted Partners

Seventeen days after the United States government forced Anthropic to shut down its two most powerful AI models for every user on earth, the standoff is showing its first signs of resolution. Anthropic has secured partial clearance for its Claude Mythos 5 model and is in active negotiations over the return of Fable 5 to general use, with reporting from Axios on June 27 indicating the Trump administration expects to restore access as soon as this week, pending final sign‑off from the Pentagon and the NSA.

The sequence of events that produced this confrontation is among the most significant regulatory interventions in the history of commercial AI.

What Happened on June 12

Anthropic launched Claude Fable 5 and Mythos 5 on June 9, 2026, to broad industry recognition. The company described Fable 5 as its most capable model for long‑horizon, agentic work — capable of running autonomously for days at a time, planning across stages, delegating to sub‑agents, and checking its own outputs. The launch included benchmark results showing state‑of‑the‑art performance across coding, knowledge work, vision, and computer use.

Three days after the launch, at 5:21 p.m. Eastern on June 12, the US government issued an export control directive requiring Anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including Anthropic's own foreign‑national employees. Because Anthropic had no clean technical mechanism to separate foreign nationals from US nationals in real time across its user base, it disabled both models for all customers globally. The API endpoint for claude‑fable‑5 returned an error. Hundreds of millions of users lost access.

Anthropic's public statement the same day disclosed its understanding of the government's concern: that a method of jailbreaking Fable 5 had been identified. The company reviewed the demonstration and disputed the characterisation. In its account, the technique amounted to asking the model to read a specific codebase and identify software flaws. The vulnerabilities surfaced were described as previously known, minor issues that other publicly available models, including OpenAI's GPT‑5.5, could produce without any bypass at all. Anthropic wrote that it had received only verbal notice of what it characterised as a narrow, non‑universal jailbreak, and argued that applying this standard across the industry would essentially halt all new model deployments by every frontier provider. A universal jailbreak, one that could broadly bypass Fable 5's safeguards, had not been demonstrated.

The Deeper Conflict

Subsequent reporting revealed that the jailbreak claim did not fully account for the speed or scope of the government's action. The context behind the directive stretches further back than June 12.

In July 2025, Anthropic had signed a deal with the Pentagon that would have made Claude the first frontier AI model approved for use on classified networks. That arrangement collapsed in early 2026 after Anthropic refused to agree to Pentagon contract terms stipulating that purchased AI models could be used for any lawful purpose. Anthropic had sought specific exclusions preventing its models from being deployed for fully autonomous weapons systems or mass domestic surveillance. The Pentagon responded in March 2026 by declaring Anthropic a supply chain risk and prohibiting defense contractors from using its models on government contracts. Anthropic filed litigation challenging that designation. A federal court found the argument plausible enough to issue a temporary block on the blacklisting while the case proceeds.

The export control directive landed three days after Fable 5's launch, during the same week Anthropic was already in active litigation against the administration. Reporting from Axios and TechCrunch indicated that Amazon CEO Andy Jassy had raised concerns about Anthropic's models with senior government officials before the directive arrived, and that the underlying security research the government cited had originated with Amazon researchers. A cybersecurity expert who reviewed that research described it as a defensive probing technique rather than an offensive jailbreak.

David Sacks, co‑chair of the President's Council of Advisers on Science and Technology, offered the administration's framing in a statement on June 14: the government had asked Anthropic to either fix the jailbreak or de‑deploy Fable 5, and Dario Amodei refused. In the administration's telling, the ban was a consequence of Anthropic's choice. Anthropic disputed this account. The NSA's position added a further dimension: NSA chief General Joshua Rudd later told the Senate that during an exercise, Mythos had breached almost all of the agency's classified systems within hours.

The Partial Resolution

The first formal step toward lifting restrictions came on June 26. Commerce Secretary Howard Lutnick sent a letter to Anthropic co‑founder and chief compute officer Tom Brown stating that appropriate safeguards were now in place to permit certain trusted partners to access Claude Mythos 5. The clearance covered approximately 100 organisations, including US government civilian agencies, national laboratories, and companies on Annex A of the authorisation document. Project Glasswing members — including Apple, Google, Cisco, Nvidia, and Microsoft — were also cleared, along with foreign‑national employees at approved organisations.

Anthropic posted on the same day that Mythos 5, described as its strongest cybersecurity model, could be redeployed to a set of US organisations that operate and defend critical infrastructure, and that the company was restoring access for those organisations quickly. The statement confirmed that negotiations were continuing to expand Mythos access and restore Fable 5 for general use.

The Commerce letter made no mention of Fable 5. All June 12 requirements remain in force for that model. Criminal and civil penalties for non‑compliance are still in effect. Lutnick reserved the right to reevaluate the approved list.

What Comes Next

As of June 28, Fable 5 remains offline pending sign‑off from the Pentagon and NSA. Other government agencies have reportedly determined that Fable 5 can safely return. Anthropic International Managing Director Chris Ciauri, speaking at the opening of Anthropic's new Korea office in Seoul, said the company was very confident that in the coming days the models would become available again. Treasury Secretary Scott Bessent and Commerce Secretary Lutnick have both been credited in reporting with helping to defuse the broader standoff.

The pricing structure for Fable 5 has changed since the suspension. As of June 23, the model is no longer included within Pro, Max, or Team plan limits and will be priced at $10 per million input tokens and $50 per million output tokens on a pay‑per‑token basis when access is restored. Anthropic updated its privacy policy, effective July 8, 2026, to include collection of government‑issued ID and biometrics for certain use cases, a mechanism that may support a future framework allowing US‑national access without fully lifting export control restrictions on foreign nationals.

The broader implications extend well beyond Anthropic. OpenAI limited the rollout of GPT‑5.6 under pressure from the Trump administration during the same week, initially releasing a preview version to a small group of government‑approved partners. The AI sovereignty debate the ban ignited internationally has been noted by policymakers from the UK to South Korea, with British minister for AI Kanishka Narayan citing the episode as an argument for deeper domestic AI investment. Legal technology firm Legion filed suit against the US government on the grounds that losing Fable 5 access caused what it described as immediate, irreparable, and existential harm to the company.

The Fable 5 episode has established, for the first time, that high‑capability AI availability is now a managed policy variable rather than a default condition. Frontier model launches will now be planned against the possibility of government intervention in ways that no AI company needed to consider before June 12, 2026.