Anthropic developed advanced AI models (Fable 5 and Mythos 5) capable of surpassing human experts in identifying and exploiting software vulnerabilities. The Trump administration issued export controls preventing foreign nationals from accessing these models, prompting Anthropic to take them offline. Over 100 cybersecurity experts and executives from major companies have petitioned the government to ease these restrictions, arguing that such controls may hinder U.S. cyber defense capabilities more than they impede adversaries, as similar AI capabilities exist internationally. The export controls represent a significant government intervention to manage AI-related national security risks. The dispute also involves concerns about the Pentagon's use of Anthropic technology and broader debates on AI risk governance.

The export controls have led to the removal of Anthropic's most advanced AI models from broader use, potentially limiting access to cutting-edge cybersecurity tools that can identify and exploit vulnerabilities. This may reduce the effectiveness of U.S. cyber defense efforts if adversaries develop or access comparable AI capabilities. The controls also reflect heightened government scrutiny of AI technologies for national security, affecting AI development and deployment strategies. There is no indication of active exploitation or direct vulnerability in the AI models themselves, but the policy impacts the availability and use of these tools.

No direct mitigation actions are required for a vulnerability or exploit since this is a policy and access control issue rather than a technical vulnerability. Organizations should monitor government advisories and policy updates regarding AI export controls. The vendor (Anthropic) has complied with the directive by taking affected models offline. Stakeholders should engage with policymakers for clarity and participate in transparent AI risk assessment processes as advocated by cybersecurity experts.