The U.S. Commerce Department has lifted export controls on Anthropic’s advanced AI models, Claude Fable 5 and Mythos 5, allowing the company to restore access starting July 1. The decision follows weeks of restrictions imposed due to security concerns, including potential vulnerabilities to jailbreaking techniques.
Anthropic confirmed the change in a post on X, stating it would begin restoring access to the models. Commerce Secretary Howard Lutnick noted in a letter that the models would no longer require export licenses, provided Anthropic adheres to security protocols and reports any malicious activity.
Background on the Restrictions
Access to the models was suspended on June 12 after the government issued export controls, blocking their use by foreign nationals, including noncitizen employees of Anthropic. The restrictions were triggered by concerns over a potential jailbreak—a method to bypass AI safeguards—though Anthropic disputed the severity of the risk, calling the vulnerabilities "relatively simple."
Anthropic’s Response
The company argued that the government’s response to the jailbreak risk was disproportionate, stating that similar vulnerabilities exist across the AI industry. It emphasized that Mythos 5 was designed with stronger cybersecurity capabilities than Fable 5, which was intended for general use.
Government’s Stance
The Commerce Department’s decision came after negotiations with Anthropic, during which the company agreed to proactively monitor security risks and collaborate with U.S. authorities on future AI model releases. The government’s initial concerns centered on preventing misuse, such as cyberattacks or other malicious activities.
Industry Implications
The reversal of restrictions highlights ongoing tensions between AI innovation and national security. Some experts argue that stringent controls could stifle progress, while others emphasize the need for safeguards against AI misuse. The case may set a precedent for how future AI models are regulated.