- Key insight: The U.S. government told Anthropic to shut down its most powerful AI model and the company complied.
- What's at stake: The episode raises questions about how much control the government has over commonly used technology.
- Forward look: Anthropic and U.S. government representatives say they want this issue resolved quickly and the AI models released back out.
In an extraordinary regulatory move, the U.S. Department of Commerce sent a letter to Anthropic directing it to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. A foreign national is a person who is not a citizen of the country they live in.
In response, Anthropic shut down all access to Fable 5 and Mythos 5.
Claude Mythos 5 is Anthropic's most advanced AI model, capable of detecting and exploiting zero-day (in other words, never before seen) software vulnerabilities at high speed. Bank CEOs and Treasury Secretary Scott Bessant
The government's letter did not provide specific details of its national security concern, according to Anthropic.
"Our understanding is that the government believes it has become aware of a method of bypassing, or 'jailbreaking' Fable 5," Anthropic said in a statement. "We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass."
David Sacks, tech investor and co-chair of the President's Council of Advisers on Science & Technology, said the government issued the order reluctantly. "It's been very surprised that Anthropic hasn't wanted to cooperate with a reasonable safety request (i.e. fixing the jailbreak issue)," he wrote on X.
According to Sacks, the government was responding to a "highly credible trusted partner" of both Anthropic and the government that was testing Fable and came forward with a jailbreak of its guardrails. The administration asked Amodei to fix the jailbreak or de-deploy the model, and Anthropic CEO Dario Amodei refused, Sacks said.
"In their blog post, Anthropic defended its decision by saying the jailbreak isn't serious," Sacks wrote. "That is not what the trusted partner and the U.S. government believe; nor is that kind of minimizing language consistent with Anthropic's brand as the AI safety company. It's difficult to fathom how they could claim a jailbreak allowing operability of a cyber weapon could be defined as not 'serious.'"
Anthropic shut down both models for all customers to comply with the directive.
The company reiterated its stance that the safeguards built into Fable "greatly reduce the likelihood that Fable is misused for tasks related to cybersecurity (among others). In fact, our safeguards are so strong that many users have complained that they are overly broad."
Anthropic also said that in the weeks leading up to the launch of Fable, it worked with the U.S. and other governments to red-team Fable's safeguards for thousands of hours.
Anthropic said it has not even received a disclosure of a concerning jailbreak that led to a harmful result. "The potential jailbreaks that have been disclosed to us are either entirely benign responses or are minor findings that provide no Mythos-specific uplift," the company said in its statement.
Anthropic apologized for the disruption to customers. "We believe this is a misunderstanding and are working to restore access as soon as possible," it said.
Sacks said the government also wants to resolve this quickly.
"The administration's hope now is that Anthropic remediates the safety issue, the export control is lifted, and Fable goes back into general release," he said. "The administration wants all of this to happen as soon as possible."
Some observers thought Amodei shot himself in the foot by saying the government should regulate AI.
"Now he got exactly what he asked for," said AI consultant Brian Rommele in an X post.
"This is straight-up leadership failure," he said. "Dario spent all that time pushing for rules and oversight. Those rules just killed his flagship models overnight. Customers in the middle of builds got cut off. Security teams using the models to find vulnerabilities suddenly had nothing. The company tried to call it a narrow export control thing over a jailbreak, but nobody is buying that spin…This whole mess was completely avoidable. Hubris dressed up as safety advocacy. "
This is not Anthropic's first run-in with the Trump Administration. In an earlier dispute with the Pentagon over the use of its models, the Pentagon ended up labeling Anthropic a supply chain risk. This made it impossible for Anthropic to obtain government contracts, but did not stop it from continuing to sell to corporations
