U.S. Lifts Anthropic AI Export Limits

Anthropic said the U.S. government has lifted export restrictions on two of its most advanced AI models, Fable 5 and Mythos 5, only weeks after the Commerce Department had ordered them restricted over national security concerns. According to Anthropic, it will begin restoring access to the models after receiving notice that the restrictions were removed. The company said the ban had been imposed because officials feared the tools could be misused by hackers to find and exploit weaknesses in computer systems. The Commerce Department, in a letter referenced by the BBC, said Anthropic had addressed the security risks tied to the models. Commerce Secretary Howard Lutnick wrote that the company agreed to proactively detect and address security risks, work with the government on future model releases, and alert authorities to malicious activity. The department also reserved the right to revisit its decision if needed. Fable and Mythos are part of Anthropic’s Claude platform, a competitor to OpenAI’s ChatGPT and Google’s Gemini. Fable 5 is described as a consumer-facing model capable of deep reasoning and complex autonomous tasks, while Mythos 5 is aimed at businesses and cybersecurity professionals and is said to be able to identify and exploit vulnerabilities in code. Anthropic previously said the government had not identified specific concerns, arguing that a suspected jailbreak attempt should not justify pulling a commercial model used by millions. The article highlights the tension between innovation, government oversight, and national security in the AI sector.

Anthropic said the U.S. Department of Commerce has lifted export controls on its Claude Fable 5 and Mythos 5 AI models, ending a tense standoff with the Trump administration that had forced the company to restrict access in mid-June. The dispute centered on national security concerns and required Anthropic to suspend model access for foreign nationals, including some of its own employees, which the company said created operational disruption. With the restriction removed, Fable 5 will again be available globally on Claude, Claude.AI, and Claude Code, while Mythos 5 has already regained limited access for some U.S. organizations and may expand further through Anthropic’s Glasswing cybersecurity program. The article frames the episode as part of a broader struggle over AI regulation in the U.S. Anthropic had rushed to Washington after the export directive, and negotiations reportedly shifted from CEO Dario Amodei to co-founder Tom Brown. Commerce Secretary Howard Lutnick said safeguards were now in place for trusted partners to access the model, while also saying the administration had worked to align the release with U.S. government priorities. The story also situates the decision within the wider race between U.S. and Chinese AI developers, noting that Anthropic’s restrictions drew criticism from tech executives and investors who feared they gave Chinese open-source competitors more time to catch up. It also compares the situation with OpenAI, which recently said it was limiting initial access to new models under government request, though it argued such access should not become a permanent norm.

The article reports that the U.S. government has lifted export controls on Anthropic’s most advanced AI models, Claude Fable 5 and Mythos 5, after weeks of negotiations over cybersecurity and model safety concerns. Anthropic said it received notice from the Commerce Department that access would be restored the next day, while Commerce Secretary Howard Lutnick said the government worked with the company to review and approve Fable 5 in order to strengthen U.S. leadership in AI. The controls had been imposed after a trusted partner—identified by CNN as Amazon—found a jailbreak that could bypass Fable’s guardrails, prompting the Commerce Department to restrict foreign-national access and causing Anthropic to disable customer access to both models. The story places this decision in the broader context of growing regulatory uncertainty around advanced AI, noting that Anthropic and the government later worked to address the cyber risks and that Mythos was allowed to be released to select government-approved entities. It also references a parallel request from the White House that OpenAI limit access to its upcoming GPT 5.6 model, underscoring a wider pattern of government caution around frontier AI systems.