Anthropic Suspends Mythos 5 and Fable 5 Models After U.S. Export Directive

The June Export Control Directive

The operational landscape for frontier AI shifted abruptly in June 2026 when the U.S. Department of Commerce issued an export control directive to Anthropic. While the directive itself was not made public, it mandated that the company suspend all access to its Mythos 5 and Fable 5 models for non-U.S. nationals, including those physically located within the United States. Anthropic, unable to reliably enforce such a restriction based on nationality, opted to disable both models for all users.

The catalyst for the measure was a security flaw identified in Fable 5. Researchers discovered that carefully crafted prompts could bypass the model’s safety classifiers, which raised concerns that the AI could be weaponized to assist in cyberattacks.

By June 30, the company had restored access to Mythos 5 for a restricted group of pre-approved U.S. government organizations and participants in Project Glasswing.

Allied Reactions and the Call for Reform

However, the incident served as a wake-up call for European policymakers. The ability of the U.S. to instantaneously cut off access to critical AI capabilities has renewed calls within Brussels for greater technological sovereignty.

AI-Driven Vulnerability Discovery

The models at the center of the controversy, Mythos 5 and Fable 5, represent a significant advancement in defensive cyber operations. These tools can analyze code with a speed and breadth that far surpasses human capability, facilitating vulnerability discovery and mitigation at an 87% success rate. In 2025, the widespread adoption of AI-driven scanning contributed to a roughly 47% reduction in mean remediation times for software flaws.

Prior to the June shutdown, the European Union had successfully negotiated an agreement with Anthropic to grant the bloc’s cybersecurity agency, ENISA, access to the highly capable Mythos 5 model. The model had previously been restricted to a select group of trusted technology companies and U.S. government agencies under Project Glasswing. As cyber leaders continue to race against adversaries, the reliability of these tools remains a critical factor in maintaining resilient digital ecosystems.

Strategic Implications for Transatlantic Tech Cooperation

The episode underscores the broader challenges of integrating frontier AI into national security frameworks. With the U.S. government acting to protect its technological edge, the need for a dedicated transatlantic forum has become more pronounced. Such a forum would allow regulators and government officials to address tensions regarding AI capabilities before they escalate into similar disruptions.

Industry leaders and policymakers are now weighing how to reconcile national security directives with the globalized nature of modern research and development. As Lindsay Gorman, Managing Director and Senior Fellow at the German Marshall Fund’s technology program, noted in discussions regarding emerging technologies, don’t let our differences get in the way of what we do share and the opportunities that we have together. The incident remains a pivot point in the ongoing debate over whether the future of cyber defense will be defined by unified transatlantic collaboration or a fragmented model of technological isolation.