Claude Mythos 5: Washington Partially Eases Blocking for Some American Companies

After several weeks of tensions around the control of advanced artificial intelligence models, the American administration has changed its position towards Anthropic. The US government now allows certain organizations to use Claude Mythos 5, while other access remains limited. This decision comes in a context where Washington seeks to strengthen oversight of sensitive technologies. The measure mainly concerns selected American partners and marks a new step in managing innovative AI tools.

In Brief

• Washington partially lifts the block on Claude Mythos 5, allowing access to more than 100 selected American companies and institutions.
• Anthropic obtains limited authorization for its AI model, while the public version Fable 5 remains suspended.
• The Department of Commerce imposes controlled access to advanced technologies, with a system based on trust and cybersecurity.
• The Glasswing project confirms the usefulness of Claude Mythos 5, notably for detecting vulnerabilities in various systems.
• The United States strengthens its oversight of artificial intelligence models, an approach that could influence the global dissemination of AI tools.

Claude Mythos 5 Regains Controlled Access with American Partners

The Department of Commerce, through Secretary Howard Lutnick, announced the end of the block targeting the export of Mythos 5 to more than 100 American institutions. This authorization concerns major companies and government agencies considered trusted partners. Claude thus benefits from a more flexible framework, but only for actors listed on an approved list by the authorities.

I have determined that appropriate safeguards are in place to allow certain trusted partners to access the Claude Mythos 5 model.

Howard Lutnick, U.S. Secretary of Commerce. Source: BE IN CRYPTO.

The decision is based on safeguards deemed sufficient by Secretary Howard Lutnick. Restrictions began after concerns about potential misuse of advanced models. Washington has maintained a cautious approach regarding the growing capabilities of artificial intelligence systems. The model was notably used in a research program aimed at identifying vulnerabilities.

Before this change, Mythos 5 participated in the Glasswing project, which brought together about 150 organizations across several countries. Tests conducted with this tool helped to detect flaws in certain sensitive systems. This experience strengthened authorities’ interest in controlled access to AI technologies. However, Anthropic continues to manage the usage conditions of the Claude model.

Washington Reinforces Its Control Over Artificial Intelligence Models

This decision comes as Washington is gradually building a new verification system for innovative models. A presidential decree has established a voluntary mechanism allowing developers to submit their creations to cybersecurity analysis before publication. The goal remains to better regulate tools capable of having a significant impact.

However, the restrictions do not only concern Anthropic. Washington has also observed similar measures with other industry players. An AI company has limited access to its most powerful model for certain authorized partners, while less advanced versions remain accessible to the public. This evolution shows a new organization of the advanced technology market.

The debate also focused on risks related to international access. Early concerns particularly involved the possibility that certain technologies could reach unauthorized actors. Cybersecurity officials have requested a review of the measures, considering that controls should remain balanced. US authorities now face different expectations depending on the companies and foreign partners.

Next Steps and Advanced AI Models

Despite the partial lifting of the block, the situation remains different depending on the versions offered by Anthropic. The public version, Fable 5, remains suspended while its development continues. Claude Mythos 5 therefore benefits from specific treatment due to its professional and institutional use.

Negotiations regarding Fable 5 remain open, and their timetable remains uncertain. This separation between professional and public models illustrates the desire to limit certain uses before wider distribution. US authorities will still need to clarify their approach in the face of the rapid evolution of the artificial intelligence sector.

Upcoming decisions could also influence relations between the United States and its allies. Several international partners are closely watching American rules concerning access to advanced technologies. For Anthropic, the authorization granted to Claude Mythos 5 represents an important step, but it does not resolve all questions related to global distribution.

The evolution of the case will therefore depend on upcoming decisions balancing security, innovation, and access to artificial intelligence models. Claude could continue to receive progressive authorizations, while US authorities will maintain their oversight of the most sensitive tools.