Cybersecurity researchers are pushing back against U.S. export controls that restrict access to Anthropic's most advanced AI models. Dozens of security experts signed a letter to the White House opposing restrictions on Fable and Mythos, Anthropic's frontier large language models.
The group argues the ban hampers domestic cybersecurity defense capabilities. Researchers need access to state-of-the-art AI tools to identify vulnerabilities, test defenses, and build safer systems. Restricting these models to only approved government entities and contractors creates an uneven playing field where private security teams operate with inferior tools compared to adversaries with unrestricted access.
The timing matters. The U.S. government has escalated AI export controls over national security concerns, particularly regarding dual-use models that could support both civilian and military applications. Anthropic's Fable and Mythos represent the company's cutting-edge offerings, putting them squarely in regulators' crosshairs. The administration's approach reflects broader anxiety about keeping advanced AI technology away from hostile nations and non-state actors.
But security researchers counter that this strategy backfires domestically. They contend that the best defense against AI-powered attacks requires access to comparable models. Cybersecurity vendors, academic institutions, and enterprise security teams all rely on frontier models to stress-test their systems and develop new protection methods. Hobbling these defenders while foreign competitors operate without such restrictions creates a perverse incentive structure.
The letter represents a rare public pushback from the security community against government policy. These experts typically work closely with federal agencies on classified projects and rarely criticize White House directives. Their willingness to publicly oppose these restrictions signals genuine concern about unintended consequences.
Anthropic has not publicly commented on the restrictions. The company previously pledged to cooperate with government oversight and has built safety considerations into its model development.