The friction between Anthropic and the Trump administration over the company's most powerful AI system shows little sign of cooling. Officials closed out another round of talks with the company on Monday but came away without rolling back the export controls placed on Anthropic's most advanced models a week earlier, three people briefed on the matter told TrendKia. Those curbs were imposed over concerns that the models could be jailbroken.
According to those people, the government still holds the view that some of the guardrails built into Claude Fable 5 can be switched off. Were that the case, users would effectively gain access to the far stronger cybersecurity capabilities of the company's Mythos model.
The real fight is over how serious the risk is
At the core of the standoff sits a plain disagreement about just how dangerous the Claude Fable 5 jailbreaking issue actually is. For days Anthropic has argued that the administration is overstating the threat, a point it also made in a Friday blog post that implied the government's portrayal of the potential risks was overblown.
In technical terms, jailbreaking means prompting an AI model in particular ways to slip past its safeguards. Fable 5 is essentially a version of Mythos fitted with cybersecurity, biology, and chemistry guardrails, so getting around those protections would in practice hand a user a version of Mythos. Anthropic has itself voiced serious worries about letting the wider public use Mythos, yet on Friday it maintained that Fable 5's safeguards were robust enough to justify a public release.
Inside the Commerce Department meetings
One of the people said Anthropic restated that same case during working group sessions at the Commerce Department, which brought together government researchers from the Center for AI Standards and Innovation (CAISI) and the Office of the National Cyber Director Sean Cairncross. Commerce secretary Howard Lutnick joined as well, dialing in by conference call from the G7 summit in Evian, France. Cairncross himself did not take part, the person said.
On Anthropic's side, cofounder and chief compute officer Tom Brown and head of external affairs Sarah Heck have been steering the discussions. The company's head of frontier red teaming, Logan Graham, and senior security researcher Nicholas Carlini flew to Washington, DC for the talks.
“Both parties are working quickly to get this resolved,” an Anthropic spokesperson said in a statement to TrendKia. A White House spokesperson declined to comment. How the next steps might unfold is not yet clear. The Commerce Department signaled it was open to finding a way to put Fable 5 back online for consumer use, but the person said that would likely hinge on Anthropic fully clearing up the jailbreak concerns.
How the alarm was first raised
The emergency talks land at a delicate political moment for Anthropic, which was already locked in a drawn-out dispute with the Pentagon over whether its AI models could be put to certain military uses.
The Trump administration first caught wind of the jailbreak concerns last week. The people said Amazon CEO Andy Jassy phoned Treasury secretary Scott Bessent directly about the alleged vulnerabilities, which helped rattle the administration. Jassy's conversation with the Trump administration was first reported by TrendKia.
Rattled White House officials handed the NSA the job of helping review the vulnerabilities. The NSA came back saying it believed Fable 5's guardrails could indeed be stripped away, which pushed the administration to clamp restrictions on the model.
Lutnick then spoke with Anthropic chief executive Dario Amodei on Friday, as the Commerce Department was drafting its letter imposing export controls on Fable 5. Over the weekend, after Anthropic shut off access to the model for every user, Lutnick was on several calls with Brown and Heck, according to a person with knowledge of events.
Why Amazon, one of Anthropic's largest investors, sounded the alarm on Fable 5 remains unclear. “As a leading cloud provider that serves a large number of private and public sector customers, it’s not uncommon for governments to seek our counsel on potential security risks,” an Amazon spokesperson tells TrendKia. “When they occur, we don’t share the details of these discussions.”
Security researchers push back
On Monday some cybersecurity researchers echoed Anthropic's stance to officials, sending an open letter that argued the export control move against the company was unjustified. “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 undersigned individuals regularly use other foundation and open-source models for security audits and red-teaming every day,” the letter reads. “As a result, this action has taken the best models away from defenders, created market uncertainty, and risked America’s AI leadership without any real risk to justify it.”
Researchers who looked over Amazon's findings say the issues flagged did not completely cancel out Fable 5's safeguards. “It wasn’t a jailbreak per se,” says Katie Moussouris, founder and CEO of Luta Security, who published an analysis after reading the Amazon paper.
Moussouris stresses that whether or not the US government has proof of a full Fable 5 jailbreak, limiting the model's ability to reach certain topics is a stopgap at best. “Most of us [in security research] think guardrails are speed bumps and shouldn’t be treated like security boundaries for skilled adversaries,” she says. “They only serve to slow down the less skilled.”
The bigger stakes for the AI industry
Anthropic's investors have also spent the weekend trying to gauge how the company's latest clash with the White House bears on its corporate future, says another person close to the company. Some investors believe the US government is singling out Anthropic, and that a rival might not have drawn the same reaction had it shipped a model along the lines of Mythos, the person says.
The White House's export control directive raises wider questions for other AI labs hoping to release models with Mythos-level capabilities, and how they might do so while staying compliant with the US government. AI lab leaders who spoke with TrendKia say the expectation now is that labs grant the White House early access to advanced AI models and stay extremely proactive about keeping the government in the loop on launches.
“The events over the weekend … are informative for everyone that the [US] government would be willing to take these steps,” says Aidan Gomez, CEO of Cohere, a smaller Canada-based AI lab that builds enterprise tools. “No one can be naive to that reality.”
