Jailbreak Panic Triggers Unprecedented AI Lockdown

Trump’s move to partially lift the Anthropic ban has reopened a fight over who controls powerful AI.

Quick Take

The Commerce Department first ordered Anthropic to block foreign access to its top two models worldwide.
The Trump administration later partially lifted that ban for a small group of American companies.
Officials said the earlier action followed a reported jailbreak that could bypass safeguards.
Critics say the government gave too little detail and too little time for a fair response.

Commerce Opens the Door, But Only a Crack

The Commerce Department has partially lifted its export ban on one of Anthropic’s most advanced models, but only for a small list of American companies. NPR reported that the new letter still keeps government control over which firms may use the model, including foreign staffers at those companies. Anthropic said it welcomed the change and would keep working with the government.^[4]^[5]

This is not a full reversal. The first order had forced Anthropic to pull both models offline for all users after the government said foreign nationals could not use them. NPR said the original ban also blocked Anthropic employees and was tied to a claim that the model’s safeguards could be bypassed. That is why the company shut the models down instead of trying to sort users by nationality.^[4]^[8]

Why the Ban Was Imposed

Officials said the crackdown came after a trusted testing partner found a jailbreak that could expose cyber capabilities in the model. One report said Commerce Secretary Howard Lutnick sent the directive to Anthropic chief executive Dario Amodei and ordered the company to make the models unavailable to any foreign national, inside or outside the United States. Supporters of the order say that kind of access risk justified a fast response.^[1]^[3]

The administration also framed the move as a national security step. A Korean report called it the first export ban on an artificial intelligence model, and other reporting said the government treated the system as a sensitive national security asset. That is a major warning sign for readers who believe Washington has gone too far in trying to police private technology, especially when the models involved are software, not hardware.^[5]

Questions Raised by the Rush Order

Anthropic said it received the directive without being given specific details of the national security concern. Other reporting said the company had only 90 minutes to comply, which left little time to test the claim or build a narrow fix. Critics argue that kind of process looks more like pressure than due process, especially when a company is told to shut down a product without seeing the full evidence.^[8]^[10]

The Commerce Department partially lifts its two-week export ban on Anthropic's most powerful model, clearing Mythos 5 for ~100 US companies and agencies while keeping Fable 5 blocked.#Anthropic #Claude

Link in the first comment 👇 pic.twitter.com/2zIiWpmwTg

— Awesome Agents (@awagents) June 27, 2026

Those doubts grew after reports that an independent cybersecurity firm viewed the alleged jailbreak as a defensive action, not a hostile one. The Wall Street Journal said the issue involved the model helping patch security flaws, and the firm called the government response an overreaction. Anthropic chief executive Dario Amodei also said the government had already decided on export controls, which fuels the sense that the policy may have been set before the facts were fully aired.^[10]^[12]

What the Partial Lift Means Next

The partial reversal does not end the fight. It shows the Trump administration is willing to keep a tight hand on frontier AI even while easing pressure for select users. That may satisfy those who want strong national security controls, but it also keeps alive the bigger question of whether Washington is creating a precedent for broad federal control over private software under vague threat claims.^[4]^[5]

For conservatives who value limited government and clear rules, the lesson is hard to miss. The federal government used emergency-style pressure, broad access limits, and a thin public record to control a private model with major economic value. The partial lift may calm some of the backlash, but it does not answer the core issue: how much power should Washington have over AI tools before the public sees the full case.^[4]^[8]