Export Ban SHOCK: Elite AI Pulled Offline

Businessperson holding a digital tablet with floating data icons — Photo: iDEAR Replay / Shutterstock

A sudden Trump administration export ban on Anthropic’s top AI tools has now been partly lifted, after the company agreed to tighten safeguards around powerful cyber capabilities that could help America’s enemies.

Story Snapshot

President Trump’s Commerce Department briefly forced Anthropic to shut down its most advanced Claude Fable 5 and Mythos 5 models worldwide over national security concerns.
Officials acted after a trusted partner reported a way to “jailbreak” the tools’ cyber features, raising fears they could supercharge hostile hacking and cyberwarfare.
The original order banned use by all foreign nationals, including Anthropic’s own non-U.S. staff, pushing the company to cut off access for every customer to stay compliant.
The Department of Commerce has now lifted export controls after Anthropic added new defenses and agreed to help the government spot and fix future security risks.

Trump Team Uses Export Powers to Lock Down High-Risk AI

The Trump administration’s move centered on two Anthropic systems, Claude Fable 5 and Mythos 5, which the company describes as its most advanced artificial intelligence tools. The United States Department of Commerce used national security export authorities to block Anthropic from providing these models to any foreign national, no matter where they lived or worked. Because Anthropic has many global customers and employees, the company decided it could not safely separate U.S. and foreign users and shut the systems down for everyone.

Anthropic’s public statement said the government order covered “any foreign national, regardless of their location, including foreign national Anthropic employees,” making selective compliance impossible in practice. This marked the first time U.S. officials had directly used export controls to stop access to a specific commercial artificial intelligence product over a claimed jailbreak with national security impact. For many conservative readers, the message was clear: under President Trump, powerful frontier technology that could be turned against America’s networks will not be left wide open.

Jailbreak Fears Tied to Cyberattack Capabilities

Officials did not release a technical report, but several outlets say the government acted after learning of a method to bypass Fable 5’s safety guards. Anthropic explained that the concern focused on a “narrow, non-universal jailbreak” that involved asking the model to read a specific codebase and find software flaws, unlocking deep cyber vulnerability scanning. Reporting links the discovery to Amazon researchers, who showed how the tool could be used to uncover bugs in software that might later be abused by a hostile attacker.

These models are known for strong cybersecurity skills and are understood to be used inside the national security community, including for offensive cyber operations. That dual-use nature—helping defenders but also potentially empowering attackers—made the alleged jailbreak a serious red flag for officials. One summary notes that a trusted partner, later identified as Amazon, alerted the government to a way around Fable’s guardrails, triggering rapid concern inside the Trump White House and the Department of Commerce. For an administration wary of foreign hacking, especially from China, taking a hard line on such a tool fits its broader national security posture.

Anthropic Pushes Back, Then Agrees to Tougher Safeguards

Even as it complied, Anthropic challenged how broad the order was. The company said it had only received “verbal evidence of a potential narrow, non-universal jailbreak” and argued that similar capabilities were already available in other public models, including competitors not placed under export controls. In plain terms, Anthropic was saying: this issue is real but small, and we do not think it justifies pulling the plug on everyone worldwide. That stance echoed industry worries that governments might overreact and harm defenders who rely on advanced tools for daily security work.

After roughly two to three weeks of shutdown, the Department of Commerce pulled back. Anthropic announced it had “received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5,” allowing it to begin restoring access. Media reports say the government’s letter credited Anthropic for taking steps to detect and address security risks and for agreeing to work with officials on standards for future models and to report possible malicious use. In effect, the Trump administration used the temporary ban as leverage to force tighter security and closer cooperation from a powerful tech vendor.

Global Allies Feel the Shock, But Security Stays Front and Center

The order did not stop at America’s borders. Because foreign nationals everywhere were covered, the ban hit users in Europe and the United Kingdom, including friendly cybersecurity teams that relied on the tools to find and fix threats. Some analysts warned the move strained alliances and fueled talk of “AI sovereignty” as other countries considered building or backing rival systems they could control themselves. This shows the trade-off conservatives have seen for years: when Washington finally clamps down on risky tech, foreign partners may complain, but U.S. leaders still have a duty to shield American networks first.

The Claude Cowork Exfiltration Vulnerability
Compounding Anthropic's security woes was the discovery of a critical, zero-day vulnerability in Claude Cowork. Independent security researchers discovered that Cowork was highly susceptible to file exfiltration via indirect prompt…

— Dodz4allai (@DrGhattasMD) July 24, 2026

Civilian cybersecurity experts also criticized the shutdown, saying it took effective tools away from defenders more than it slowed down attackers who can grab other models. At the same time, the quick partial lifting of controls once Anthropic improved safeguards shows the policy was not meant as a permanent chokehold, but as an emergency brake while a serious concern was checked. For conservative readers, this episode highlights both the promise and the peril of advanced artificial intelligence: in the wrong hands, frontier models can become weapons, and a White House that moves fast to lock them down is acting squarely in defense of American security and sovereignty.