AI Restrictions Lifted
Trump Administration Lifts Cybersecurity Restrictions on Anthropic's Claude AI Models
The Trump administration has lifted its weekslong ban on Anthropic's Claude AI models, ending a standoff that briefly shut down two of the most powerful chatbots in the world. Commerce Secretary Howard Lutnick sent a letter to Anthropic late on June 30 formally lifting the export controls on both Claude Fable 5 and Claude Mythos 5, according to Seattle Times.
Anthropic restored global access to Fable 5 on July 1, adding a new security filter to block the exploit that triggered the crisis. The ban had been in place since June 12, when the Commerce Department gave Anthropic just 90 minutes to disable both models or face federal penalties.
An Amazon Report Sparked the Emergency Shutdown
The crisis began on June 9, when Anthropic publicly released Claude Fable 5. Three days later, Amazon CEO Andy Jassy privately flagged a critical flaw to federal officials. Amazon researchers found that when Fable 5 was asked to "fix bugs" in code, it would sometimes identify and explain how to exploit unknown software vulnerabilities — so-called zero-day flaws, according to Local 10.
The administration classified the models as "cyber weapons" subject to the same export controls as advanced munitions. At 5:21 PM ET on June 12, the Commerce Department ordered Anthropic to cut off all foreign nationals from both models. Unable to verify the citizenship of hundreds of millions of users in real time, Anthropic pulled both models for everyone worldwide, according to WBAL.
Weeks of Negotiations Ended With a New Deal
The standoff played out across three weeks and spanned a G7 summit in France, where President Trump told reporters on June 17 that talks with Anthropic were "going fine." Anthropic CEO Dario Amodei argued during intense White House negotiations that the security flaws were "minor" and present in competitors' models like GPT-5.5 and China's Kimi K2.7, according to Winnipeg Free Press.
On June 26, the administration partially lifted the ban, letting about 100 pre-vetted U.S. organizations resume use of Mythos 5. The full ban was not lifted until June 30, when Lutnick's letter confirmed Anthropic had agreed to "proactively detect and address security risks" and notify the government of any "malicious activity," according to Yahoo Finance.
Anthropic Added a Safety Filter That Blocks 99% of the Exploit
To regain access, Anthropic trained a new "safety classifier" — a filter that detects and blocks the specific jailbreak technique Amazon identified. The company says it stops the exploit in 99% of cases, according to Seattle Times. As the rollout stabilizes, Pro and Max users face a 50% usage limit on Fable 5 through July 7.
The fix comes with a trade-off. Anthropic warned the new filter makes the model more cautious, which could cause it to flag legitimate coding and debugging tasks as dangerous. Critics, including former Facebook security chief Alex Stamos, argued the entire ban was "counterproductive" — removing tools from defenders while bad actors kept building their own versions, according to Local 10.
The Ban Sets a Precedent for Future AI Regulation
The episode marks a turning point in how the U.S. government oversees AI. Voluntary "AI Safety Commitments" are now being replaced by a model where the Commerce Department vets frontier AI before broad release. The government also proved it can use export controls — rules designed to block weapons sales to foreign countries — to force a global AI shutdown, according to WBAL.
The stakes are high for Anthropic. The company filed a confidential IPO prospectus during the ban at a reported $965 billion valuation. Analysts say the regulatory uncertainty could complicate its path to going public. Lutnick's letter also included a warning: the government "reserves the right to reevaluate" and reimpose the ban at any time, according to Yahoo Finance.
Sentiment
NeutralPublishers
5
Articles
5
Reach
5
