Anthropic moves toward deal with US to lift AI model curbs

Anthropic is in active negotiations with the Trump administration to restore international access to its most advanced AI models, after a government directive earlier this month effectively cut the company off from global markets. The dispute has pulled in Nvidia, Adobe, and the US Department of Defense along the way.

How we got here

On June 12 and 13, 2026, the Trump administration directed Anthropic to suspend access to its AI models for foreign nationals. The order came from the US Department of Commerce, driven by national security concerns that had two distinct threads.

The first involved SK Telecom’s alleged connections to China. The second was a reported jailbreak of the guardrails built into Fable 5, attributed to Amazon.

Anthropic complied. The company disabled worldwide access to both Fable 5 and Mythos 5, its two flagship models, effectively locking out every international user and partner overnight.

On June 15, 2026, executives from Nvidia and Adobe sent a joint letter to the US government urging a reversal. Their core argument: Fable 5 and Mythos 5 are not uniquely suited for offensive military applications, and restricting them does more damage to American cybersecurity than it prevents. Cybersecurity professionals, the letter argued, depend on these tools for defense, not attack.

A meeting between Anthropic’s technical team and Commerce Department officials followed on June 16, 2026.

The Pentagon shadow hanging over this

Earlier in January 2026, a clash with the Department of Defense came to a head over the application of Anthropic’s AI systems in mass surveillance and fully automated weapons systems. Anthropic opposed both use cases, firmly and publicly. The result was the cancellation of a Pentagon contract worth approximately $200 million.

What this means for investors and the AI industry

The involvement of Nvidia and Adobe in the June 15 letter is notable precisely because neither company is a disinterested party. Nvidia sells the chips that train and run these models. Adobe integrates AI capabilities into its creative suite. Both companies have a direct commercial stake in an environment where AI models can move freely across borders.

The reported jailbreak of Fable 5’s guardrails also raises a separate investor consideration. If Anthropic’s safety architecture can be circumvented, and that circumvention becomes the justification for government intervention, then every AI company’s security posture becomes a material business risk. Regulatory exposure is no longer just about what your model does. It’s about what someone else can make your model do.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.