US lifts export restrictions on Anthropic’s most powerful AI models

The US Department of Commerce completely lifted export restrictions on Anthropic’s flagship models, Claude Fable 5 and Claude Mythos 5, on June 30, 2026, The New York Times reports. This decision allows the company to restore global access to its most advanced artificial intelligence developments, which had been unavailable to users worldwide for nearly three weeks.

Ban on all foreigners, including company employees

The incident began on June 12, 2026, when the agency issued an official export control directive to Anthropic. The company was ordered to immediately suspend access to the Fable 5 and Mythos 5 models for all foreign nationals—regardless of whether they were outside or inside the US, including its own foreign employees. Unable to verify users' nationality in real time, Anthropic decided to shut down the models completely for all customers to avoid violations, Reuters reports.

According to Anthropic's official statement published on its website, national security concerns prompted this drastic step. The government did not disclose specific details, but sources link it to discovered vulnerabilities that could theoretically allow users to bypass built-in guardrails and use the AI for malicious actions, including cyberattacks. Fable 5 was positioned as one of the most powerful publicly available models, while Mythos 5 was an even higher-performing version with relaxed restrictions for trusted cybersecurity partners.

National security threat or unfounded panic?

Bloomberg notes that the shutdown caused significant inconvenience for users and businesses. Anthropic quickly dispatched specialists to Washington and began intensive negotiations with the administration. As a result, on June 30, Commerce Secretary Howard Lutnick sent a letter to the company stating that following a risk assessment, the requirement for export, re-export, and technology transfer licenses was lifted. The company confirmed that access restoration began the very next day.

Threat classifier strengthened

By the time the models returned, Anthropic had implemented additional security measures, DW News reports. Specifically, a new threat classifier has been introduced, which blocks suspicious requests in over 99% of cases. According to Anthropic’s official materials, the new safety classifier triggers on prompts related to cybersecurity (exploit development, vulnerability scanning and exploitation, malware creation, reconnaissance, lateral movement, and other stages of cyberattacks), biology and chemistry (research potentially involving harmful substances), as well as attempts at model distillation to train competing AI systems. When triggered, the request is not completely blocked but is automatically redirected to the less powerful Claude Opus 4.8 model, and the user receives a corresponding notification. The company acknowledges that the classifier may be overly strict on programming and debugging tasks, calling it a "work in progress." Additionally, temporary usage caps (50% of normal limits) remain in effect until July 7.

A precedent for AI regulation

As The Wall Street Journal writes, the conflict has become a notable precedent in AI regulation. This is one of the first instances where advanced artificial intelligence models were effectively treated as controlled technologies under export control rules. Experts note that such government intervention demonstrates the US administration's desire to maintain US technological leadership amid global competition. Anthropic also proposed an industry framework for assessing the severity of jailbreak vulnerabilities in collaboration with partners (Amazon, Microsoft, Google, and others).

In parallel with resolving the Anthropic situation, similar measures were discussed regarding other companies, such as OpenAI, which, as Al Jazeera and NPR report, agreed to additional screening of users for its new models. The incident affected not only commercial clients but also foreign partners, including companies and organizations from allied nations.

Despite the swift resolution, the incident demonstrated the industry’s vulnerability to regulatory risks. The department's letter notes that Anthropic has committed to proactively identifying and mitigating threats, but the authorities reserve the right to reinstate controls in the event of violations.