Anthropic AI Export Controls Lifted
U.S. Government Lifts Export Ban on Anthropic's Powerful Mythos and Fable AI Models
Anonymous
The Trump administration lifted all export controls on Anthropic's Fable 5 and Mythos 5 AI models on June 30, ending a 15-day global blackout that had cut off customers worldwide. Washington Post reported the White House announcement came at 9:07 PM ET, with Anthropic beginning the phased restoration of access to both models the following day.
The ban had forced Anthropic to disable both models entirely on June 13, just one day after the U.S. Commerce Department issued an emergency export control order. The company said it had no reliable way to verify the nationality of API users in real time, so it cut access for everyone, UPI reported.
A Jailbreak Claim Set Off the Sudden Ban
The Commerce Department issued its emergency order on June 12, citing a reported "jailbreak" of Fable 5. The government alleged that the model's safety guardrails could be bypassed, allowing it to operate at the more powerful Mythos level — a tier designed for cybersecurity defense. Officials said the model could identify vulnerabilities in critical infrastructure, making it a "dual-use weapon" that needed strict oversight.
Anthropic pushed back hard. CEO Dario Amodei called the ban "improvised" and "opaque," arguing the so-called vulnerabilities were already findable using other public AI models. He said "perfect jailbreak resistance is not currently possible" for any AI system. According to Forbes, Amazon CEO Andy Jassy was allegedly the person who first shared the jailbreak discovery with the government, triggering the June 12 order.
Amodei Confronted Trump Directly at G7 Summit
The dispute went all the way to the top. On June 17, Dario Amodei confronted President Trump about the ban during a working lunch at the G7 Summit in Evian-les-Bains, France, according to Washington Post. Amodei reportedly also had tense calls with the White House and the National Cyber Director in the days that followed.
The administration held firm at first. On June 26, Commerce Secretary Howard Lutnick sent a letter to Anthropic partially reversing the order — but only for about 100 "trusted" U.S. institutions and government agencies, Miami Herald reported. Lutnick said the models required "alignment across the US Government" before full access could be restored.
15-Day Blackout Left Global Customers Scrambling
Before the ban, Mythos 5 had access in 200 institutions across 15 countries. All of them lost access overnight. The blackout forced enterprise teams to rethink how much they depended on a single AI provider. Anthropic, which recently hit a $47 billion annual revenue run rate — up from $10 billion the year before — now faces hard questions about its reliability as a global provider, according to AI Weekly.
Bipartisan pressure also mounted in Congress. Rep. Sam Liccardo (D-CA) and Rep. Jay Obernolte (R-CA) led a group of lawmakers demanding transparency on the legal basis for the ban. Sen. Angus King (I-ME) questioned whether the order was a real security measure or the result of what he called "announced antipathy" between the administration and the company, according to San Luis Obispo Tribune.
Ban Sets a Precedent for All Future Frontier AI Models
The episode is already reshaping how the AI industry operates. Analysts at Goldman Sachs described the event as proof that frontier AI is becoming "state-supervised strategic infrastructure" — no longer just a commercial product. Under Executive Order 14409, future frontier models from OpenAI, Meta, and others are now expected to go through a 30-day voluntary government review before public deployment.
European leaders at the G7 expressed alarm that their economies now depend on U.S. AI tools that can be "withdrawn without warning," according to Washington Post. OpenAI CEO Sam Altman said he was uncomfortable with "the government picking the customers," even as OpenAI agreed to a similar pre-screening process for its own GPT-5.6 model. The 15-day Anthropic blackout has accelerated calls in the EU and UK for "sovereign AI" to cut reliance on U.S.-controlled models.
Sentiment
NeutralNEGATIVENEUTRALPOSITIVE
Publishers
6
Articles
4
Reach
6
Topics
A
AnonymousAuthor
