The administration believes there are ways to disable some guardrails on Anthropic’s Cloud Fable 5, effectively allowing users to access the more powerful cybersecurity capabilities of the company’s Mythos model, the people said.
Anthropic has said for days that the administration’s concerns are overblown, a position reiterated in working group meetings held with government researchers from the Center for AI Standards and Innovation at the Commerce Department and the office of national cyber director Sean Cairncross, one of the people said.
The meetings also included Commerce Secretary Howard Lutnick, who spoke via conference call from the G7 summit in Evian, France. Cairncross himself did not participate, the person said.
On the Anthropic side, co-founder and chief compute officer Tom Brown and head of external affairs Sarah Heck are leading the discussion. Anthropic’s head of frontier red-teaming, Logan Graham, and senior security researcher Nicholas Carlini flew to Washington, DC for 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 what next steps might be taken. The Commerce Department has expressed a willingness to find a way to bring the Fable 5 back online for consumer use, the person said, but that will likely depend on Anthropic fully resolving the jailbreak concerns.
the alarm is ringing
The emergency talks come at a difficult political moment for Anthropic, which was already in a protracted battle with the Pentagon over whether its AI models could be used for certain military applications.
The Trump administration was first alerted to the jailbreak concerns last week. Amazon CEO Andy Jassy called Treasury Secretary Scott Besant directly about the alleged vulnerabilities, which played a role in scaring the administration, the people said. Jesse’s conversation with the Trump administration was first reported by The Information.
Concerned White House officials tasked the NSA with helping review the vulnerabilities. The NSA responded that it believed it was indeed possible to remove the Fable 5’s handrail, prompting the administration to ban the model.
Lutnick spoke with Anthropic Chief Executive Dario Amodei on Friday as the Commerce Department prepared its letter to impose export controls on the Fable 5. Over the weekend, Lutnick was on several calls with Brown and Heck after Anthropic closed access to the model to all users, according to a person with knowledge of the events.
It’s unclear why Amazon, one of the biggest investors in Anthropic, sounded the alarm over the Fable 5. “As a leading cloud provider that serves a large number of private and public sector customers, it is not unusual for governments to seek our advice on potential security risks,” an Amazon spokesperson told WIRED. “We do not share the details of these discussions when they occur.”
safety disconnect
At the core of the talks between Anthropic and the administration is a disagreement over the severity of CloudFable 5 jailbreaking concerns.
In a blog post on Friday, Anthropic said the administration’s characterizations of potential risks were exaggerated. Some cybersecurity researchers reiterated this position to authorities on Monday, sending an open letter arguing that the export control action taken against Anthropic was unfair.
The open letter reads, “Anthropic’s Mythos-class models are quite good at finding flaws and weaponizing exploits. However, they are not uniquely good at these tasks, and many of the individuals signed below regularly use other foundation and open-source models for security audits and red-teaming.” “As a result, this action has taken the best models away from the defenders, created uncertainty in the market, and put America’s AI leadership at risk without any real risk to justify it.”
<a href