Anthropic's Fable 5 and Mythos 5 Back Online After US Export Control Standoff

Regulatory Truce Ends 18-Day Export Control on Anthropic AI Models

The United States Department of Commerce has officially lifted export controls on Anthropic's advanced artificial intelligence models, Claude Fable 5 and Mythos 5. This decision, confirmed by US Commerce Secretary Howard Lutnick on July 1, 2026, marks the end of an 18-day period during which access to the models was restricted globally. Full access to Fable 5 is expected to be restored worldwide from July 1.

The restrictions, which were imposed on June 12—just three days after Fable 5's launch—followed a directive from the Trump administration. This action came after the government learned of a report by Amazon researchers detailing a method to bypass Fable 5’s integrated safeguards. Anthropic, unable to reliably verify user nationality in real time, made the decision to temporarily suspend global access to both Fable 5 and Mythos 5. The company noted that the initial government directive did not provide an explanation of the underlying national security rationale for the controls.

Collaborative Resolution and Enhanced Safeguards

Over the subsequent weeks, Anthropic engaged in close collaboration with the US government and key partners, including Amazon, to address the identified vulnerabilities. This extensive review process led to the development and training of an improved safety classifier, which the company states is now capable of blocking the reported bypass technique in over 99% of cases. Commerce Secretary Lutnick emphasized the collaborative effort, stating that the work aimed to ensure alignment across the US Government and reinforce America’s leadership in AI.

Beyond resolving the immediate issue, Anthropic is actively participating in a broader initiative with other industry leaders such as Amazon, Microsoft, and Google, as part of the Glasswing partners group. This collaboration seeks to establish a comprehensive framework for assessing and mitigating AI "jailbreaks" – techniques used to circumvent model safety features.

Wider Market Impact and Future Regulatory Landscape

The resolution of the export control issue has had a tangible impact on the AI sector. Following the lifting of restrictions, AI chip stocks, including Micron and Intel, collectively saw an addition of $2 trillion in market value, signaling renewed investor confidence. In a related development, Amazon Web Services (AWS) has announced a significant commitment of $1 billion towards establishing a dedicated AI engineering unit, further underscoring the industry's push into advanced AI capabilities.

This incident vividly illustrates the evolving landscape of AI regulation, particularly concerning "dual-use technologies" that have both civilian and military applications. Governments worldwide are increasingly classifying advanced AI models, especially large frontier models capable of sensitive tasks, under export control regulations like the US Export Administration Regulations (EAR). This means that API access to AI models can be restricted, shut down, or geofenced with little warning, posing a significant operational risk for enterprise AI builders who invest substantial resources into integrating these models into their workflows. While the Biden administration's January 2025 AI Diffusion Rule, which segmented global access to AI technology, was slated for revision by the Trump administration, the overarching trend points towards tightening controls on advanced AI, not loosening them.

The rapid imposition and subsequent lifting of controls on Anthropic's models highlight the critical balance governments seek between fostering innovation and managing potential national security risks in the fast-paced AI industry.