Trump Administration Orders Anthropic to Block Foreign AI Access

The U.S. government has ordered Anthropic to cut off foreign access to its most advanced AI systems, intensifying a months-long conflict between the San Francisco-based company and the Trump administration.

The directive came over the weekend, days after Anthropic released two new models: Fable 5, designed for software engineering and scientific research, and Mythos 5, a cybersecurity-focused tool released only to select defenders and infrastructure providers. In response, Anthropic disabled access to its most advanced models for all customers while maintaining availability of other systems.

The security concerns

According to Anthropic, the government cited national security issues but provided limited specifics. The company stated in a Friday blog post that officials believe a method exists to bypass Fable 5's safeguards—a technique known as "jailbreaking." Anthropic maintains that protective measures remain in place and that no tester has successfully circumvented these controls at scale.

The Los Angeles Times reported that conversations between Amazon CEO Andy Jassy and Trump administration officials, including Treasury Secretary Scott Bessent, triggered the presidential order. Amazon researchers reportedly demonstrated that Fable 5 could provide information useful to cyberattackers. Amazon, which has invested billions in Anthropic and supplies the company with training chips, told the Wall Street Journal that governments routinely seek its input on potential security risks.

David Sacks, who served as the administration's crypto and AI advisor, accused Anthropic of prioritizing product availability over safety. Defense Secretary Pete Hegseth referenced an earlier decision to remove Anthropic from a Department of Defense building, stating that recent events validate that choice.

Why it matters

This dispute highlights the growing tension between AI companies racing to deploy powerful models and government efforts to control national security risks. Anthropic's compliance—disabling access for hundreds of millions of users globally—sets a precedent that could reshape how frontier AI models are released. The company warns that applying this standard industry-wide would "essentially halt all new model deployments for all frontier model providers," potentially slowing AI development at a critical juncture when Anthropic is competing for dominance and preparing for a public offering at a nearly $1 trillion valuation.

A pattern of conflict

This marks the second major clash between Anthropic and the Trump administration. Earlier disputes centered on a $200 million Pentagon contract, with Anthropic seeking guarantees against using its technology for domestic surveillance or autonomous weapons. In February, Trump directed federal agencies to stop using Anthropic's technology, labeling the company "radical left" and "woke." Anthropic subsequently sued the administration over its designation as a "supply chain risk," alleging illegal retaliation.

Next steps

Senior Anthropic technical staff traveled to Washington on Monday for meetings with White House officials at the Commerce Department, according to sources familiar with the discussions. Both parties are working to address the security concerns, though the timeline for restoring access remains unclear.

These details were first reported by the Los Angeles Times.