White House – The Trump administration says it wants Anthropic to address alleged vulnerabilities in its advanced Claude model after export controls took the model offline. But cybersecurity experts and even the administration’s own resource limits raise doubts about whethe
The Trump administration’s fight with Anthropic is no longer just about what the company claims is being overblown. It is about what the government expects to be fixed before the most advanced model comes back out into the world.
After Trump officials took Anthropic’s Claude Fable 5 offline with export controls last week over concerns about jailbreaking—using prompts to get around a model’s safeguards—the administration is pressing the company to take concrete steps. The goal. officials say. is simple: if Anthropic wants to rerelease Claude Fable 5. it needs to address what the government alleges are vulnerabilities.
Anthropic has been arguing for days that the administration’s worries are overstated and that the effects of the jailbreaks are minimal. The company reiterated that position to the Commerce Department and the Office of the National Cyber Director. Sean Cairncross. during a technical meeting on Monday.
But government officials say the argument over how serious the jailbreaks are has started to narrow. They are now focused on the National Security Agency’s conclusion that there are ways to disable guardrails on Fable 5—the safeguards put in place to prevent users from accessing capabilities of the Mythos model related to cybersecurity. chemistry. and biology.
At this stage, the administration essentially treats the problem as something Anthropic must solve. Three people familiar with discussions say the expectation is that the company will need to be more proactive about testing not only Fable 5. but all of its frontier AI models. to find potential jailbreaks and flag them to the government itself.
The administration’s reasoning is also practical. Neither the Commerce Department’s Center for AI Standards and Innovation nor the National Security Agency. people familiar with the discussions say. has the staff or bandwidth to chase down every conceivable jailbreak on every model that reaches the market.
There is a deeper snag beneath all of that. Independent cybersecurity experts have increasingly taken the view that guardrails on AI models are only a stopgap solution. Skilled users—and future AI models—can find ways to bypass constraints. That view leaves the administration’s demand looking harder to fulfill than it sounds.
A White House spokesperson declined to comment.
The dispute over jailbreaks is unfolding as other parts of the administration’s intelligence agenda face their own disruption. At the start of the week. Trump’s pick to serve as Acting Director of National Intelligence. Bill Pulte. was on track to never even start the job. Trump then threw him a lifeline with a change that also shifted the pressure onto the permanent DNI nomination.
The permanent DNI nominee, Jay Clayton, now faces the prospect of never serving in the role. Trump initially named Pulte—his housing finance chief—to replace outgoing DNI Tulsi Gabbard. But Trump faced bipartisan pushback because Pulte does not have the national security experience required by law for the role. and because he flagged allegedly questionable mortgage fraud accusations against Trump’s political enemies.
Trump responded by announcing Clayton, the US attorney for the Southern District of New York, as his nominee for a permanent DNI. Gabbard was scheduled to depart June 18, with Pulte’s first day set for June 19.
Senate Republicans questioned whether the timeline could hold. They wondered whether Clayton could have his hearing fast-tracked to June 17 and start by June 22—whether Pulte would even get into the building. On Wednesday, Trump scrapped that plan.
As part of a wider feud with Senate Republican leadership over the filibuster. Trump announced Clayton’s hearing would be delayed indefinitely in an apparent effort to prevent Pulte from getting jumped. Senate Republicans then said the hearing would proceed unless Clayton didn’t appear or his nomination was withdrawn.
The political turmoil may be a body blow for the Office of the Director of National Intelligence. Trump has directed Pulte to vastly downsize the office. and staffers have been unimpressed by what they see as Pulte’s minimal effort to get to know the agency and lack of regular briefings. people familiar with the matter said.
Anthropic Claude Fable 5 jailbreaking export controls guardrails cybersecurity Office of the National Cyber Director Sean Cairncross National Security Agency Center for AI Standards and Innovation Bill Pulte Jay Clayton Tulsi Gabbard Acting Director of National Intelligence DNI
So they shut it down… because people might jailbreak it. Makes sense?
I don’t get why this is a “press” thing, like just tell Anthropic what to fix. Also export controls took it offline? So who cares now, it’s already off the shelf.
This is gonna sound dumb but isn’t a jailbreak just like hacking the app? If the NSA can “disable guardrails” then… doesn’t that mean they already know how to get around it anyway? Seems kinda convenient.
“Address vulnerabilities” is such a vague phrase. Like, how many times have we heard that from tech companies. And “frontier AI models”?? there’s like 10 different ones in the story already. Also export controls = politics more than cybersecurity, IMO. If they really cared, they wouldn’t be arguing about how serious jailbreaks are for days.