Anthropic has regained permission to release its most advanced AI models globally after strengthening safeguards that the U.S. government had restricted for national security reasons. The San Francisco-based AI company said on Tuesday that the Department of Commerce lifted export controls on Claude Fable 5 and Mythos 5, models that were blocked from foreign access on June 12 after researchers discovered vulnerabilities in their safety systems.

Why Did the U.S. Government Ban These AI Models?

The Trump administration ordered Anthropic to immediately restrict access to Fable 5 and Mythos 5 just three days after Fable 5 launched on June 9. Washington feared that foreign adversaries could exploit these powerful AI systems to mount devastating cyber-attacks. The concern became urgent after Amazon researchers published a report showing how to bypass Fable 5's safeguards to make it identify software vulnerabilities, a capability that could be weaponized by malicious actors.

The ban represented a significant hardening of the Trump administration's stance toward AI regulation. Previously, the White House had emphasized a lighter regulatory touch, prioritizing speed in developing AI technology to maintain competitive advantage over China. However, national security officials began treating frontier AI models, which are the most advanced systems available, with the same caution applied to nuclear weapons.

"In conversations with many of the president's other national security and economic security advisers, we're talking about the impact of these frontier AI models. It would be not misplaced to refer to their capabilities as akin to digital nuclear weapons," said CIA Director John Ratcliffe during a speech at the AWS summit in Washington.

John Ratcliffe, CIA Director

What Changed to Allow the Models Back Online?

Anthropic worked closely with the U.S. government over the past two weeks to strengthen safeguards around Fable 5 and Mythos 5. Commerce Secretary Howard Lutnick confirmed in a letter to the company that "Anthropic has worked with the US government to address risks associated with the Covered Models". The company increased its safety measures and agreed to several commitments going forward.

Researchers from the U.S. Department of Commerce's Center for AI Standards and Innovation tested both Anthropic's previous and new safeguards. According to Anthropic, these researchers agreed that the updated safeguards are "extraordinarily strong". The company also committed to proactively detecting and addressing security risks, working with the government on protocols for future releases, and reporting any malicious activity to federal authorities.

How to Understand the New Government-AI Company Framework

• Voluntary Testing Framework: The Trump administration issued an executive order on June 2 calling for federal agencies to create a voluntary framework for private companies like Anthropic and OpenAI to test and release powerful frontier AI models in collaboration with government oversight.
• Restricted Initial Access: Before full public release, advanced models now receive government vetting to determine which organizations can access them, a practice that has drawn criticism from some AI leaders.
• Ongoing Coordination: Anthropic said it would "scale up our government collaboration" and establish dedicated teams to work on shared government priorities, signaling a new era of public-private partnership in AI development.

The restoration of access to Fable 5 and Mythos 5 marks a partial reversal of the June 12 ban, though the process remains contentious. OpenAI, which also delayed its GPT-5.6 model release at the government's request, has expressed reservations about the approach. OpenAI CEO Sam Altman stated that while extensive safety testing "is not a bad idea," he dislikes "the idea of the government picking the customers".

The broader context reflects growing concern among intelligence agencies about frontier AI capabilities. Last week, intelligence agencies from Australia, the United States, the United Kingdom, New Zealand, and Canada, collectively known as Five Eyes, warned that while AI would improve cyber defense over time, it also accelerates the speed, scale, and sophistication of cyber threats. These agencies emphasized that the timeline for transformative AI capabilities is "not years, it is months".

United Kingdom, New Zealand, and Canada, collectively known as Five Eyes

Susie Wiles, the president's chief of staff, praised the cooperation from technology companies in implementing the administration's AI and cybersecurity executive order. She highlighted "excellent work around advanced model access and guardrail testing and security," framing the government-industry collaboration as unprecedented and foundational to an "America First" approach.

She

Anthropic said it would begin restoring global access to Fable 5 and Mythos 5 on Wednesday, July 2, ending the more than two-week blackout that had prevented foreign nationals from using the models anywhere in the world.