U.S. Eases Restriction, Allows Anthropic Mythos 5 Access for Select Agencies and Companies
U.S. Commerce Department permits controlled redeployment of Anthropic Mythos 5 to over 100 government and private organizations, restoring access under specified safeguards.
The U.S. government has moved to restore limited access to Anthropic Mythos 5, reversing parts of a recent restriction that removed the model from general availability. The decision allows the model to be used by more than 100 designated U.S. government agencies and private companies, and it explicitly permits non-U.S. employees at those organizations to access the model. The announcement follows a period of intense scrutiny after researchers demonstrated ways to bypass the model’s safety guardrails.
U.S. Decision to Redeploy Mythos 5
On June 26, 2026, Commerce Secretary Howard Lutnick informed Anthropic in a letter that safeguards were considered sufficient to permit targeted redeployment of the Claude Mythos 5 model. The letter authorized access for a defined set of trusted partners that operate and defend critical infrastructure. The move reverses parts of an earlier ban that had removed Mythos 5 from the market and barred many non-American users.
The administration framed the change as narrowly tailored, intended to balance operational needs for powerful cybersecurity tools with national security concerns. Officials said the decision followed close review of Anthropic’s mitigation measures and operational controls for how the model would be hosted and used.
Scope of Access and Eligible Organizations
The redeployment applies to slightly more than 100 specific U.S. government agencies and private-sector companies deemed critical to national infrastructure. Eligible users include both American and non-American employees who work for those designated organizations, restoring access that had been restricted under the prior directive. Anthropic’s own non-American engineers who support the model are also included among those now permitted to use Mythos 5.
Access will be limited to organizations that satisfy security and oversight requirements outlined by the Commerce Department. The restrictions are intended to confine the model’s use to operational activities such as cyber defense, incident response, and protective services for sectors considered vital to national security.
Government Rationale and Required Safeguards
The Commerce Department cited a determination that appropriate technical and procedural safeguards are in place to reduce risks posed by high-capability models. Those safeguards reportedly include usage monitoring, strict access controls, and provisions for human oversight in sensitive tasks. The government indicated that these measures were a prerequisite for authorizing access to organizations responsible for critical infrastructure defense.
Officials emphasized the decision does not return Mythos 5 to general availability and that the authorization can be revised if safeguards prove insufficient. The government signaled ongoing review and potential expansion or contraction of permitted access based on results from the controlled redeployment.
Fable 5 Excluded from Redeployment Directive
The recent directive did not address the wider release of Fable 5, a variant of Mythos 5 that was distributed publicly days before the initial ban. Fable 5 had been presented by Anthropic as a more protective build, and it remained available for public use prior to government intervention. The absence of Fable 5 in the redeployment order leaves its future availability unresolved.
Anthropic had pulled both Mythos 5 and Fable 5 from general distribution after researchers demonstrated that the models’ guardrails could be bypassed. The government’s selective restoration of Mythos 5 underscores a distinction in how the two models are being evaluated and managed. Observers noted that a separate determination will be needed before Fable 5 can be re-released or cleared for broader use.
Anthropic’s Public Statement and Ongoing Talks
Anthropic publicly acknowledged progress in restoring access and said it had been working closely with the U.S. government since June 12, 2026. In a social media statement, the company said Mythos 5 would be redeployed quickly to the designated organizations and that it continued to discuss expanding access and making Fable 5 available again. Anthropic indicated ongoing collaboration with regulators to address outstanding safety concerns and operational requirements.
The company’s statement framed the redeployment as a step toward restoring broader availability, but it also signaled that additional work is required to satisfy regulatory expectations. Anthropic did not disclose technical specifics tied to the safeguards but said deployment would proceed under the conditions set by government reviewers.
Security Risks and Industry Implications
Security researchers and some industry stakeholders have warned that high-capability models like Mythos 5 raise distinct risks if control measures are inadequate, including potential misuse in cyberattacks or the generation of sophisticated exploit code. The government’s decision to restrict access to trusted partners reflects those concerns and represents a novel approach to balancing innovation with national security. Experts said the controlled redeployment will serve as a real-world test of whether technical and organizational controls can mitigate misuse.
The selective restoration may influence how other companies develop and harden advanced AI models intended for defensive use. Observers expect companies to accelerate investments in monitoring, access governance, and red-team testing to demonstrate that strong guardrails can survive adversarial probing. The outcome of this phased deployment is likely to shape future policymaking on the export, distribution, and oversight of powerful AI systems.
The government and Anthropic have both signaled that the situation remains fluid, with further adjustments possible as the controlled rollout proceeds and additional technical evaluations are completed.
