The U.S. Commerce Department has removed an export restriction that had prevented Anthropic from making its Claude Fable 5 and Mythos 5 artificial intelligence models available outside the United States. Anthropic announced the decision on X and thanked users for their patience.
What happened and when
In mid-June Anthropic suspended access to Fable 5 and Mythos 5 for users located outside the United States in order to comply with U.S. export controls. The company said it was asked by the government to suspend all foreign access, including for foreign nationals who were physically present within the United States.
Officials cited tests indicating the technology could be highly dangerous, capable of breaching designated security systems in a short time; reporting also noted that Anthropic's systems performed rapidly even against systems linked to the NSA.
Access restored and the Glasswing program
According to CNBC, the two models became available again on July 1 via the Claude platform, including Claude.AI and Claude Code. Anthropic also said it restored Mythos 5 access for certain U.S. organizations after the government approved the program on June 26.
Anthropic and the U.S. government said they continue to work together to extend Mythos 5 access to additional domestic and international partners. The Glasswing program is intended to provide selected organizations controlled access to advanced AI models for testing security systems.
Government actions, political context and security concerns
CNBC's reporting notes that the government's action against Anthropic coincides with the rapid rise of Chinese open-source models, which are becoming nearly as capable as the strongest U.S. models while being significantly cheaper. Restrictions imposed under the Trump administration on Anthropic's model rollouts prompted concern among technology leaders and investors that Chinese developers could gain valuable time to catch up.
Relations between Anthropic and the U.S. government have previously been strained. The Trump administration banned certain government entities from using the company's models after Anthropic refused permission for military use, and Trump publicly criticized the company. Despite political tensions, reporting indicates the Pentagon used the system in planning scenarios related to Iran.
Representation in negotiations
In talks with the government, Anthropic was represented by co‑founder Tom Brown, replacing CEO Dario Amodei in that role. The sources note that Dario Amodei publicly supported Kamala Harris in the 2024 presidential election.
Why this matters
Lifting the restriction allows Anthropic's advanced models to become available more broadly, albeit under controlled conditions and ongoing government oversight focused on safety guarantees. The case underscores the tension between advanced AI capabilities and national security, and highlights how geopolitical and market competition — particularly the growth of Chinese open-source models — can influence regulatory decisions.
