Washington re-authorizes the use of Anthropic's most advanced AI for a limited group of partners

The government has once again authorized access to Anthropic’s most advanced artificial intelligence model, Mythos 5, though only for a small group of U.S. partners involved in cybersecurity and the protection of critical infrastructure, according to the company.

The decision comes two weeks after the U.S. government ordered access to the model to be blocked for all foreign users on national security grounds. According to Anthropic, the restoration of service will initially benefit a limited group of “cyber defenders and infrastructure operators” in the United States, while it continues to work on expanding access.

For the time being, international organizations and partners, including government cybersecurity agencies in Europe and Asia, remain without access to the platform.

Likewise, the situation regarding Fable 5, the version of Mythos 5 intended for the general public and designed with limitations on capabilities related to cybersecurity and biological or chemical risks, remains unresolved. The company noted that it is in discussions with the government to expand access to the advanced model and restore the availability of Fable 5.

Blocked for security reasons

On June 12, Secretary of Commerce Howard Lutnick ordered the suspension of access by foreign nationals to both models after security vulnerabilities were detected. According to published reports, the decision came after Amazon reported flaws in Fable 5’s security systems.

In a letter cited by various media outlets, Lutnick stated that Anthropic has worked alongside the government to reduce the risks associated with these models, which has reportedly allowed for sufficient progress to partially authorize their reactivation.

For his part, a spokesman for the Department of Commerce, Benno Kass, stated that the administration seeks to maintain U.S. leadership in artificial intelligence without compromising national security.

Greater oversight of advanced models

The partial reactivation of Mythos 5 coincided with the launch of GPT-5.6, OpenAI’s new model, access to which would also be subject to an individual validation process by the U.S. government, according to available information.

OpenAI CEO Sam Altman stated that the procedure does not represent the ideal model for the industry, although he praised the administration’s handling of the challenges posed by this technology.

These decisions reflect a shift in the policy of President Donald Trump’s administration regarding the development of artificial intelligence. After years of advocating for limited regulation to promote competitiveness against China, the government signed an executive order in early June establishing a federal review process for advanced models prior to their commercial release.

However, the proposed mechanism is voluntary and non-binding.