In a dramatic policy shift that took the tech world by surprise, the Trump administration on July 4 lifted federal restrictions on Anthropic’s most advanced artificial intelligence systems, clearing the way for the company to immediately reactivate its flagship models. The decision de-escalates a months-long standoff between the White House and the San Francisco-based AI firm, which had accused regulators of stifling innovation under vague national security pretexts.
Sources familiar with the matter confirm that the Office of Science and Technology Policy issued a revised directive late Thursday, removing Anthropic’s Claude 4 and Claude 5 architectures from the restricted export and domestic deployment list. The models, which some experts describe as possessing “frontier-level reasoning capabilities,” were effectively frozen in mid-May after the Department of Commerce flagged them for potential dual-use risks. Anthropic had argued the ban was arbitrary and lacked evidence of any specific threat.
The reversal comes just weeks after Anthropic CEO Dario Amodei publicly warned that the restrictions would hand a strategic advantage to Chinese AI labs. He testified before Congress on June 18 that the ban had already forced his company to halt key research partnerships with U.S. defense contractors. “This is a victory for American technological leadership,” Amodei said in a statement Friday. “We can now resume full operations and deliver these tools to customers under robust safety protocols.”
Critics, however, are raising alarms. The Electronic Privacy Information Center (EPIC) filed an emergency petition with the Federal Trade Commission on July 5, arguing that the sudden greenlight bypassed standard interagency risk reviews. “We are concerned that political pressure, not science, drove this decision,” said EPIC president Caitriona Fitzgerald. “These models can generate synthetic biological agents and manipulate financial markets at scale. The public deserves transparency.”
Industry analysts note that the timing is politically charged. With midterm elections looming in November, President Trump has been touting a “Made in the USA AI agenda” on the campaign trail. The White House press office did not respond to requests for comment, but a senior official speaking on condition of anonymity told Reuters that the administration “determined that existing voluntary commitments from Anthropic are sufficient to manage risk.”
Meanwhile, Anthropic has already begun rolling out the Claude 5 API to enterprise clients in healthcare and logistics. The company says it will publish a third-party audit of the model’s behavior by August 1. For now, the AI landscape has shifted overnight—and the debate over how fast America should run with its most powerful technologies is far from settled.