The Anthropic AI export ban has come to an end, as the Trump administration reversed a hastily imposed restriction that had blocked the company from offering its most advanced models to customers worldwide. The decision restores global access to Anthropic’s powerful Fable and Mythos systems after more than two weeks of uncertainty.

The reversal marks a notable resolution to a standoff that had rattled the tech industry and raised thorny questions about national security and AI governance.

The End of a Contentious Ban

For over two weeks, the U.S. government effectively cut off access to Anthropic’s Fable and Mythos AI models. The restrictions stemmed from national security concerns, driven largely by the models’ ability to identify software vulnerabilities, a capability that alarmed officials worried about cybersecurity risks.

Commerce Secretary Howard Lutnick announced the shift on Tuesday, saying in a post on X that the government had spent the previous two weeks working closely with Anthropic to analyze and approve the Fable 5 model. He framed the effort as a way to ensure alignment across the U.S. government while strengthening America’s leadership in AI.

Anthropic confirmed the news in its own social media post, noting that the Commerce Department had informed the company the ban was being lifted on both the Fable and Mythos models.

The Terms of the Agreement

The reversal came with strings attached. According to a letter from Lutnick to Anthropic reviewed by The Washington Post, the company agreed to proactively detect and address security risks associated with its technology.

The letter also indicated that Anthropic would continue working with the U.S. government to develop standards for future model releases, signaling an ongoing collaborative relationship between the company and federal regulators.

In a blog post late Tuesday, Anthropic said it had modified the Fable 5 model to make it even stricter in refusing cybersecurity-related requests, an apparent effort to address the very concerns that had prompted the ban in the first place.

How the Restrictions Unfolded

The timeline reveals just how quickly the situation escalated. Anthropic had previously agreed to let the government approve new customers of Mythos, but export controls imposed on June 12, which barred access to non-U.S. citizens, ultimately led the company to shut off access entirely.

Fable, which offers similar capabilities but with more robust safety controls, had previously been available to the public. The dual restrictions left one of America’s leading AI companies unable to serve much of its global customer base.

Anthropic’s rival OpenAI faces related scrutiny. The company is still required to obtain government approval for new users of its own top model, GPT-5.6, which was announced just last week.

A Shift in the White House Approach

The episode reflects a striking evolution in how the Trump administration handles AI regulation. The tight controls over whom Anthropic and OpenAI can serve represent a sharp departure from the laissez-faire philosophy Trump championed upon returning to office last year.

Just days after his inauguration, Trump rescinded Biden-era AI rules, declaring that he wanted to unshackle the tech industry so it could innovate rapidly and outpace China in a global AI arms race. Yet in recent months, as worries about the cybersecurity implications of AI have intensified, the White House has reinvented itself as a far more assertive regulator.

Industry Frustration and Global Pushback

The ban on Fable and Mythos did not sit well with parts of the tech world. Some industry leaders expressed anger, frustrated that they could not use the technology to enhance their own companies and products during the shutdown.

The ripple effects extended overseas as well. Foreign leaders cautioned that restrictions on the global use of American AI could leave their countries at a disadvantage, and some called for individual nations to build their own technology free from U.S. interference.

For now, the lifting of the ban resolves the immediate crisis and allows Anthropic to resume serving its worldwide customers. Yet the underlying tensions between rapid innovation, national security, and global access remain very much unresolved, suggesting that this latest chapter is unlikely to be the last in the debate over how to govern powerful AI.