In a swift reversal of a weeks-old national security order, the U.S. Department of Commerce has removed export restrictions on two of Anthropic’s most cutting-edge artificial intelligence models, the AI developer has confirmed. The company announced Tuesday it will restore global access to its consumer-facing Claude Fable 5 and enterprise-focused Mythos 5 starting Wednesday, just 21 days after it abruptly halted access to both tools at the U.S. government’s request. Launched only three days before the June 12 suspension, the two models represent Anthropic’s most advanced AI work to date, positioning the startup as a leading competitor to OpenAI’s ChatGPT and Google’s Gemini. The restrictions were implemented over federal concerns that the powerful models could be exploited by malicious actors to compromise digital systems, with specific focus on a potential ‘jailbreak’ loophole that could bypass built-in safety controls to unlock unrestricted capabilities. Mythos 5, built for business and cybersecurity clients, is specifically designed to detect and test exploits for coding vulnerabilities, which raised red flags for national security officials. In an official letter to Anthropic reviewed by the BBC, Commerce Secretary Howard Lutnick confirmed that the company has fully addressed the identified risks. Under the new agreement reached with regulators, Anthropic has committed to proactive ongoing monitoring to detect and mitigate security threats linked to both models. The firm will also coordinate with federal authorities on all future AI model releases and immediately report any suspicious or malicious activity related to its platforms. Lutnick emphasized that the Commerce Department retains full authority to reimpose restrictions if future risks emerge that are not adequately mitigated by Anthropic’s new safeguards. At the time of the initial suspension, Anthropic pushed back on the government’s action, noting that regulators had not outlined specific, detailed flaws in its technology beyond the broad report of a potential narrow jailbreak. The company contested the decision to pull a commercial product already deployed to hundreds of millions of users over an isolated, limited vulnerability. Fable 5, Anthropic’s flagship consumer model, is engineered for advanced deep reasoning and can execute complex, multi-step tasks independently, while Mythos 5 caters to professional cybersecurity teams supporting digital infrastructure defense. The BBC has reached out to the U.S. Department of Commerce for additional comment on the lifted ban and the terms of the new regulatory agreement.