How the Trump-Anthropic Feud Escalated to an Export Control Block on Frontier AI

Anthropic received the order on a Friday afternoon with no advance warning and, according to the company, "did not specify the national security concerns" behind the action .

The Roots of the Conflict: A Six-Month Escalation

The June export control was not an isolated event. It was the decisive escalation in a bitter, months-long confrontation centered on a single question: who draws the ethical red lines around frontier AI ?

January 2026: The Pentagon's Demand and Anthropic's Refusal

In late January, Reuters reported that the Pentagon was demanding Anthropic remove safety guardrails from its Claude models so the military could use them for autonomous weapons targeting and domestic surveillance under a contract potentially worth up to $200 million . Anthropic refused, insisting on keeping its usage policies intact. The dispute reached an impasse after several weeks of negotiations .

February 27, 2026: The Administration Strikes Back

On February 27, President Trump announced on Truth Social that he was directing all federal agencies to "IMMEDIATELY CEASE all use of Anthropic's technology," declaring, "We don't need, we don't want, and will not do business with them again!" . That same day, the Pentagon formally designated Anthropic a "supply-chain risk" and placed the company on a national security threat blacklist—a label historically reserved for foreign adversaries like Huawei . Defense Secretary Pete Hegseth also threatened to invoke the Cold War-era Defense Production Act to force Anthropic to hand over Claude for military use. Anthropic stated it would go to court if the administration followed through . Meanwhile, the Pentagon finalized an alternative AI agreement with OpenAI .

April 2026: A Brief Pause

By late April, Politico reported that the administration was "backing off" from its most extreme threats. Despite the supposed federal ban, some government offices, including the Commerce Department's own Center for AI Standards and Innovation, were already testing Mythos . The underlying mistrust between Anthropic and the administration, however, remained unresolved.

June 12, 2026: The Export Control Hammer

The June 12 order was the administration's sharpest escalation. The Commerce Department reportedly acted after another company claimed it could bypass or "jailbreak" a safeguard designed to prevent Fable 5 from being used to detect software vulnerabilities . With the Pentagon already blacklisting Anthropic as too hazardous for government use, the Commerce Department's licensing framework now deemed the company too dangerous for international access .

Anthropic's Response: Full Compliance, Sharp Criticism

Anthropic complied immediately but made its objections unmistakably clear. Because it could not practically distinguish foreign from domestic users in real time, the company took both Fable 5 and Mythos 5 completely offline for all customers worldwide.

"The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance," the company said in a statement . Access to all other Anthropic models, including the widely used Claude chatbot, was not affected .

Anthropic publicly stated it "disagrees with the government's handling of the matter" and noted the directive lacked specific national security details . The company called for a transparent statutory licensing framework, arguing that "the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts" .

The order also created an extraordinary internal situation: Anthropic's own foreign-national employees were locked out of the models they had built. As one industry observer noted, "The munition is in the building and the people who made it are not allowed to look at it" .

The Wider Landscape: Microsoft, xAI, and Policy Shifts

While the Trump administration escalated its confrontation with Anthropic, several parallel developments reshaped the competitive and regulatory landscape.

Microsoft Moves to Cut Anthropic Out

At its Build 2026 conference (June 2–5), Microsoft unveiled seven in-house "MAI" AI models and signaled a strategic shift to reduce dependence on both OpenAI and Anthropic . Microsoft AI chief Mustafa Suleyman publicly stated the company's goal is to "reduce and ultimately eliminate" payments to Anthropic, calling Claude's costs too expensive and pitching Microsoft's own MAI-Thinking-1 as a replacement . This was a significant blow given Microsoft's role as a major enterprise distribution partner for Anthropic.

xAI, Google, and Microsoft Join Pre-Release Review

On May 5, 2026, Bloomberg reported that xAI, Google, and Microsoft had agreed to give the U.S. government early access to their AI models for pre-release security evaluation by the Center for AI Standards and Innovation (CAISI) . OpenAI and Anthropic had already joined similar voluntary arrangements. This created a jarring contradiction: Anthropic was participating in the government's voluntary safety review system while simultaneously being hit with unilateral, post-deployment export controls on those same models.

Anthropic's Policy Evolution Amid the Storm

Even as it defended its safety red lines, Anthropic was evolving its policy positions. In May 2026, the company published "2028: Two scenarios for global AI leadership," arguing democracies must maintain a lead in advanced AI and supporting strict chip export controls on China—a stance broadly aligned with the administration's stated goals . By early June, Anthropic's policy page emphasized support for pre-release testing, independent evaluations, and safety-incident disclosure . And in a notable corporate move, Anthropic filed a confidential S-1 with the SEC around this period, signaling preparations for a potential IPO despite the hostile regulatory climate .