Anthropic gets global green light for its most advanced models after passing safety tests

What happened?

The U.S. government has lifted export restrictions on Anthropic's most advanced artificial intelligence models, Claude Fable 5 and Claude Mythos 5. The decision, announced on June 30, 2026, comes just three weeks after the Trump administration ordered the suspension of these models for being considered a national security risk.

In a letter to Anthropic cited by Reuters and The New York Times, Commerce Secretary Howard Lutnick stated that Anthropic "will no longer need a license for exports or domestic transfers of its Claude Mythos and Claude Fable models." The letter acknowledges that the company "has taken steps in close coordination with the U.S. government to address the risks" posed by these systems.

As a result, Claude Fable 5 is now available worldwide, while access to Mythos 5 has been restored for U.S. organizations since June 26. Anthropic is also working with the government to expand access to Mythos through the Glasswing program, which allows cybersecurity researchers at trusted companies to use the model for defensive purposes.

Why is this important?

This episode marks a milestone in the regulation of advanced artificial intelligence. For the first time, a government has subjected a frontier AI model to a national security review process, imposed restrictions, and then lifted them after verifying that adequate mitigation measures have been implemented. This sets a precedent for future interactions between AI developers and governments, suggesting that collaboration, rather than confrontation, may be the way forward.

The decision also has geopolitical implications. By lifting restrictions, the U.S. allows Anthropic to compete globally with other players such as OpenAI, Google DeepMind, and Chinese companies, which are also developing frontier models. Anthropic's ability to operate without export licenses simplifies its international expansion and accelerates the adoption of its technologies.

What will be the consequences?

In the short term, users and companies worldwide will be able to access Claude Fable 5, which according to Anthropic offers significant improvements in reasoning, creativity, and language understanding. Mythos 5, meanwhile, remains a high-performance model restricted to defensive cybersecurity uses, but its availability will gradually expand.

In the long term, this case could serve as a model for other countries seeking to balance AI innovation with national security. The transparency and cooperation demonstrated by Anthropic could become an industry standard. Additionally, the existence of the Glasswing program suggests that governments are willing to create controlled channels for advanced AI to be used for security benefits, provided technical and contractual safeguards are in place.

However, questions remain about the sustainability of this approach. Can smaller startups meet the same security requirements? Will government approval become a barrier to entry for new players? How will open-source models without a responsible corporate entity be handled? These questions remain open.

What should readers know?

• The Claude Fable 5 and Mythos 5 models were halted in June 2026 by the Trump administration due to national security concerns, related to their potential use in cyberattacks or disinformation.
• Anthropic collaborated with the government to implement security measures, such as usage restrictions, continuous monitoring, and participation in the Glasswing program.
• The approval is not a blank check: Mythos 5 remains subject to access controls, and any misuse could reverse the decision.
• This case sets a precedent for frontier AI regulation, where cooperation between company and government can resolve security conflicts without stifling innovation.

"The decision to lift restrictions shows that it is possible to find a balance between technological advancement and national security, as long as there is transparency and willingness to collaborate." — Analyst at TheVortiq.