The Trump administration is reversing export controls on Anthropic's two most advanced AI models, Mythos and Fable, according to Wired. The decision comes just weeks after the White House ordered the company to suspend access for foreign users — a reversal that underscores the turbulent and at times contradictory AI policy that has defined the administration in 2026.

Imposed and lifted in record time

In June 2026, the Department of Commerce introduced licensing requirements for the export of Anthropic's Mythos 5 and Fable 5 to all foreign users, according to background information cited by Wired. The justification was concern over "jailbreak" vulnerabilities and potential national security risks associated with highly capable AI systems. At the same time, OpenAI was pressured to restrict access to its GPT-5.6 models to government-approved entities.

Now the administration is easing those very same restrictions — without, as of yet, providing a detailed public explanation for the rapid about-face.

The administration has consistently argued that overly strict export controls could "stifle American innovation" and undermine the United States' long-term AI leadership.
Trump reverses course: Lifts export controls on Anthropic's most powerful AI models - Bilde 1

A policy in constant flux

This reversal is not an isolated incident. In May 2025, the administration repealed Biden's so-called AI Diffusion Rule, which would have introduced worldwide licensing requirements for advanced AI models and computing chips. Under Secretary of Commerce Jeffrey Kessler characterized Biden's policy at the time as "poorly conceived and counterproductive," and promised a more open approach toward allied nations — combined with strict controls against rivals.

Balancing security and market access

The shifting policy reflects a fundamental tension: the administration wants to preserve the United States' technological edge in the AI race — particularly against China — while fearing that overly harsh restrictions will damage American companies and push countries toward non-American alternatives.

China is the central sticking point. U.S. export controls on advanced AI chips from companies such as NVIDIA, AMD, and Intel, as well as manufacturing equipment from ASML and Applied Materials, have been in place since Trump's first term and have since been tightened further. The rationale is China's "military-civil fusion" strategy, which blurs the line between civilian and military technological development.

Frontier AI models are now treated as potentially as sensitive an export commodity as advanced semiconductors — but enforcement is far more difficult.

Uncertainty around frontier AI as an export commodity

The swift reversal on the Anthropic restrictions raises questions about how carefully considered the original decision was. Export control of software models is in principle far more complex than controlling physical hardware: model weights can be distributed digitally, and licensing mechanisms are difficult to enforce consistently on a global scale.

Wired reports that the White House eased the restrictions without providing a comprehensive public explanation. It has not yet been confirmed whether the relief applies fully to all foreign users, or whether geographic carve-outs will be introduced — for example, for countries deemed strategic adversaries.

Regardless, the administration's handling of the Anthropic case illustrates a broader pattern: U.S. AI policy in 2026 is being shaped on an ongoing, ad hoc basis, under heavy influence from a technological landscape that is evolving faster than regulation can keep pace with.