The United States has lifted export restrictions on Anthropic’s most advanced artificial intelligence models, ending a weeks-long block that had prevented public access to the company’s Mythos and Fable systems and intensified debate over how Washington should regulate powerful AI technology.
Anthropic said Tuesday that the US Commerce Department had withdrawn the requirement for the company to obtain a license before making the models available abroad. The company said it would begin restoring access to the systems on Wednesday, July 1, following the removal of export controls imposed less than three weeks earlier.
The restrictions were introduced on June 12, when the US government added Anthropic’s models to its list of export-controlled technologies. The decision meant the systems could not be provided to foreign nationals without special approval. Anthropic said complying with the requirement at a large scale was impractical, leading the company to suspend public access to the models entirely.
The move followed growing government scrutiny of advanced AI systems over concerns that they could be misused by foreign governments, including military or intelligence agencies in countries such as China and Russia. US officials have sought greater oversight of frontier AI models, arguing that the technology’s capabilities could create new national security risks if released without adequate safeguards.
In a letter to Anthropic, US Commerce Secretary Howard Lutnick said the export controls were being withdrawn after the company agreed to several commitments. Lutnick said Anthropic had agreed to proactively identify and address security risks associated with its models, work with the US government on protocols and standards for Mythos, Fable and future systems, and notify the government of malicious activity involving the technology.
Anthropic said it had received notice from the Commerce Department that export controls on Claude Fable 5 and Mythos 5 had been lifted and that access would begin being restored. The company had previously committed publicly to many of the same security measures before the export restrictions were introduced.
The reversal came after weeks of uncertainty surrounding the future availability of Anthropic’s models. Mythos was initially released in April to a limited group of organizations as the company attempted to address concerns about its ability to identify and exploit software vulnerabilities. A public version of Fable was released in June with additional security protections.
The restrictions had also raised questions about the US approach to AI regulation. Some cybersecurity experts argued that the export controls did not directly address the security concerns surrounding the models and instead created uncertainty for AI developers. The decision also came as US technology companies faced increasing competition from overseas AI developers producing systems with similar capabilities.
Asian AI companies have begun releasing models approaching the capabilities of Anthropic’s systems, including Fugu and Tulonfeng, increasing pressure on US officials to ensure American AI companies remain competitive internationally.
Before the restrictions were lifted, the US government had already allowed limited access to Mythos for selected customers approved by the White House. Rival AI company OpenAI also faced restrictions around the release of its latest models, with access limited to a small group of vetted partners instead of a full public launch.
The Trump administration’s approach to AI oversight has created uncertainty across the industry. An executive order signed in June established a voluntary framework encouraging developers of advanced AI models to provide systems to the US government for review for up to 30 days before releasing them to approved partners. The proposal drew criticism from some analysts who questioned how future model releases would be governed.
Isaac Harris, executive director of the Frontier Security Institute, said the decision to lift the restrictions indicated that a process for setting standards around US AI models was developing. However, he also raised questions about how the administration would address similarly advanced AI capabilities developed overseas with fewer safeguards.
Lutnick said the Commerce Department could reconsider the decision if circumstances changed or if Anthropic failed to meet its commitments. He said the government reserved the right to restore a license requirement if necessary.
Both Mythos and Fable are based on the same underlying AI model, although Fable is designed for broader public availability while Mythos includes fewer restrictions. Anthropic has also introduced a new safeguard targeting a technique the company believes the government viewed as a method for bypassing protections, known as jailbreaking. A company source said the vulnerabilities associated with that technique had already been identified and patched.
Security experts have warned that powerful AI systems such as Mythos could accelerate sophisticated cyberattacks if misused, particularly against sectors such as banking that rely on complex and aging technology infrastructure.
The dispute adds to a difficult period in Anthropic’s relationship with the US government. Earlier this year, the Pentagon designated the company a supply-chain risk, preventing contractors working for the US military from using Anthropic’s AI after the company declined to allow its models to be used for mass domestic surveillance or fully autonomous weapons systems.
With the export restrictions removed, Anthropic’s models are again available as the company and US officials continue negotiating the balance between expanding AI innovation, maintaining global competitiveness and addressing security concerns surrounding increasingly capable artificial intelligence systems.
