Trump administration officials concluded negotiations with Anthropic on Monday without removing the shipping controls that were in place posted last week more than a company advanced models of AI in response to jailbreak concerns, according to three people briefed on the matter.
The administration continues to believe there are ways to disable some of the layers of protection on Anthropic’s Claude Fable 5, allowing users to access the greater cybersecurity capabilities of the company’s Mythos model, the people said.
Anthropic has said for many days that the administration’s concerns are overblown, a position that was emphasized in meetings of a working group held at the Department of Commerce and government researchers from the Center for AI Standards and Innovation and the Office of the National Director of the Internet, Sean Cairncross, one of the people said.
The meetings were also attended by Commerce Secretary Howard Lutnick, who joined by conference call from the G7 summit in Evian, France. Cairncross himself was not involved, the person said.
For Anthropic, co-founder and chief computing officer Tom Brown and chief external affairs officer Sarah Heck have been leading the discussion. The head of Anthropic’s red team, Logan Graham, and senior security researcher Nicholas Carlini flew to Washington, DC, for the talks.
“Both parties are working quickly to resolve this,” an Anthropic spokesperson said in a statement to WIRED. A White House spokesman declined to comment.
It was not immediately clear how any of the next steps might play out. The Commerce Department expressed interest in finding a way to bring Fable 5 back online for consumer use, but it may be up to Anthropic to fully resolve the jailbreak concerns, the person said.
Ringing the Bell
The emergency talks come at a difficult political time for Anthropic, which was already long overdue fight the Pentagon on whether its AI designs could be used for certain military applications.
The Trump administration was first notified of the jailbreak concerns last week. Amazon CEO Andy Jassy called Treasury Secretary Scott Bessent directly about the alleged vulnerability, which played a role in scaring the administration, the people said. Jassy’s conversation with the Trump administration was the first information and News.
Frustrated White House officials tasked the NSA with helping review the vulnerability. The NSA responded that it believed it was indeed possible to remove Fable 5’s security guards, prompting the administration to put restrictions on that model.
Lutnick then spoke with Anthropic chief executive Dario Amodei on Friday, when the Commerce Department produced its letter imposing a shipment control on the Fable 5. Over the weekend, after Anthropic cut off the model’s availability to all users, Lutnick was on the phone with Brown and Heck, according to a person with knowledge of the incident.
It’s unclear why Amazon, one of Anthropic’s largest investors, sounded the alarm on Fable 5. “As a leading cloud provider serving a large number of private and public sector customers, it’s not uncommon for governments to seek our advice on potential security risks,” an Amazon spokesperson tells WIRED. “As they occur, we do not share details of these discussions.”
Disconnect Security
At the heart of the negotiations between Anthropic and the administration is a disagreement over the severity of the Claude Fable 5 jailbreak problems.
In a blog post on Friday, Anthropic indicated that the administration’s characterization of potential risks has been overstated. Some cyber security researchers reiterated this position to officials on Monday, sending an email open letter arguing that the export control measures taken against Anthropic were unfounded.
“Anthropic Mythos-class models are very good at finding flaws and exploiting weapons. However, they are not particularly good at these tasks, and many signatories regularly use other core and open-source models for daily security checks and red teaming,” the open letter reads. “As a result, this move has removed the best examples from advocates, created market uncertainty, and jeopardized America’s AI leadership without any risk of justification.”
