White House Forces Anthropic to Disable AI Models Over Security Fears

The Trump administration has taken the extraordinary step of forcing Anthropic to shut down its most powerful AI models, marking the first time the federal government has deemed a leading AI company's technology too risky for public use.

The action came swiftly after Amazon CEO Andy Jassy alerted White House officials on Friday that researchers had discovered ways to bypass safety guardrails in Anthropic's Fable 5 model, potentially enabling cyberattacks. President Trump agreed with the assessment, and his team immediately imposed foreign export restrictions on the company. To comply, Anthropic disabled both Fable 5 and its even more advanced Mythos 5 system—products the company had positioned as among the most capable AI tools on the market.

According to The Wall Street Journal, which first reported the details, Anthropic has been in negotiations with the administration since the weekend attempting to restore user access.

Why it matters

This shutdown signals a fundamental shift in how the U.S. government will regulate AI development. After months of taking a hands-off approach to preserve American competitiveness against China, the administration is now asserting direct authority to evaluate and potentially block AI releases based on national security concerns. For AI companies racing to deploy increasingly powerful systems, the precedent is clear: federal officials can now shut down your flagship product within hours if they deem it a security risk.

Industry pushback and broader concerns

The move has sparked immediate backlash from the AI research community. Dozens of prominent researchers and technology workers signed an open letter criticizing the action, arguing it "has taken the best models away from defenders, created market uncertainty and risked America's A.I. leadership without any real risk to justify it."

Anthropichas defended its systems by noting that other models, including some open-source alternatives, possess similar capabilities. Several cybersecurity experts who reviewed the Amazon findings told the Journal that other models can perform comparable functions and characterized the government's response as an overreaction.

The clash comes against a backdrop of existing tension between Anthropic and federal agencies. The Defense Department classified Anthropic as a supply chain security risk several months ago following disputes over Pentagon AI use and appropriate safeguards, leading to multiple lawsuits.

A new regulatory framework emerges

President Trump recently signed an executive order expanding the role of national security and cybersecurity officials in evaluating AI models before public release. The Anthropic shutdown represents the first major test of this framework.

Amrith Ramkumar, tech policy reporter at The Wall Street Journal who has covered the situation closely, noted that the incident reflects growing government concern about advanced AI capabilities. "These latest models are no joke," he reported officials and researchers saying. "They're capable of cyberattacks. They could be capable of biological weapons and all kinds of other stuff without the right guardrails."

The administration now faces the challenge of balancing security concerns against maintaining American AI leadership, particularly as Chinese companies and government entities develop their own advanced systems.

This reporting was first detailed by Amrith Ramkumar of The Wall Street Journal and discussed in a PBS NewsHour segment.