Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Trump administration officials concluded talks with Anthropic on Monday without lifting export controls that had been in place imposed last week On most of the company Advanced artificial intelligence models In response to jailbreak concerns, according to three people familiar with the matter.
Management still believes there are ways to disable some of the guardrails in Anthropic’s Claude Fable 5, effectively allowing users to access the company’s Mythos model’s more robust cybersecurity capabilities, the people said.
Anthropic has said for days that the administration’s concerns are overblown, a position it reiterated in working group meetings held at the Commerce Department with government researchers from the Center for Artificial Intelligence Standards and Innovation (CAISI) and the office of National Cyber Director Sean Cairncross, one of the people said.
Also attending the meetings were Commerce Secretary Howard Lutnick, who called in via conference call from the G7 summit in Evian, France. Cairncross himself was not involved, the person said.
On the Anthropic side, discussions were led by co-founder and chief computing officer Tom Brown and head of external affairs Sarah Hick. Anthropic’s Frontier Red Team Lead, Logan Graham, and Senior Security Researcher Nicholas Carlini traveled to Washington, D.C., for the talks.
“Both parties are working quickly to resolve this issue,” an Anthropic spokesperson said in a statement to WIRED. A White House spokesman declined to comment.
It was not immediately clear how the next steps would proceed. The Commerce Department has expressed a willingness to find a way to bring Fable 5 back online for consumer use, but that would likely be conditional on Anthropic fully resolving jailbreak concerns, this person said.
Alarm ringing
The emergency talks came at a fraught political moment for the Anthropists, which was already long in the making Fighting with the Pentagon On whether its AI models could be used in certain military applications.
The Trump administration was first alerted to the prison break concerns last week. Amazon CEO Andy Jassy contacted Treasury Secretary Scott Piscent directly about the alleged vulnerabilities, which played a role in raising the administration’s concerns, the people said. Jassy’s conversation with the Trump administration was his first I mentioned By information.
Panicked White House officials assigned the National Security Agency to help review the vulnerabilities. The NSA responded that it believed it was indeed possible to remove Fable 5’s guardrails, prompting the administration to impose restrictions on the model.
Lutnick then spoke with Anthropic CEO Dario Amodei on Friday, as the Commerce Department drafted its letter imposing export controls on Fable 5. Over the weekend, after Anthropic cut off access to the model for all users, Lutnick made several calls to Brown and Heck, according to a person familiar with the events.
It’s unclear why Amazon, one of Anthropic’s largest investors, has sounded the alarm about Fable 5. “As a leading cloud provider serving a large number of private and public sector customers, it is not uncommon for governments to seek our advice on potential security risks,” an Amazon spokesperson told WIRED. “When they happen, we don’t share the details of those discussions.”
Disconnect security contact
At the heart of talks between Anthropic and management is a disagreement over the severity of Claude’s Fable 5 jailbreak concerns.
In a Blog post On Friday, Anthropic noted that management’s characterization of the potential risks was exaggerated. Some cybersecurity researchers reiterated that position to officials on Monday, sending a letter Open letter Under the pretext that the export control measures taken against Anthropic are unjustified.
“Anthropic’s Mythos class models are very good at finding flaws and weaponizing vulnerabilities. However, they are not uniquely good at these tasks, and many of the undersigned individuals regularly use other core and open source models for security audits and red teaming every day,” the open letter said. “As a result, this action took the best models away from advocates, created uncertainty in the market, and risked America’s leadership in AI without any real risk to justify it.”
