What to know about the Anthropic models takedown

Anthropic took down its newest AI models last week, after the Trump administration slapped the company with an export control order requiring it to block foreign nationals from accessing its latest technology.

The order comes amid concerns about the potential capability to bypass the models’ guardrails, sparking yet another dispute between Anthropic and the administration.

Here’s what to know about the situation:

Anthropic pulls Fable, Mythos models over ‘jailbreak’ concerns

Just days after releasing its Fable 5 and Mythos 5 models, Anthropic announced Friday that it was removing access to both in response to the administration’s directive, which cited national security concerns.

The pair are based on the company’s new, more powerful Mythos model. It initially declined to offer Mythos to the public amid concerns about boosting hacking capabilities, instead providing limited access to certain companies and government agencies.

Anthropic opted to release Fable 5 last week with guardrails in place meant to prevent potentially dangerous uses. Mythos 5, which was offered to a smaller contingent of cyber-defenders and infrastructure providers, had fewer safeguards than its counterpart.

The concerns about a potential “jailbreak” of these guardrails were reportedly first raised by Amazon CEO Andy Jassy. Amazon has invested in Anthropic, most recently committing $5 billion to the AI startup on top of an earlier $8 billion.

The administration pushed Anthropic to voluntarily pull the models before hitting it with the export control order, according to several outlets.

White House, Anthropic disagree over extent of threat

Anthropic’s resistance to taking down its models stems from a disagreement with the White House over the extent of the threat.

In a letter to the company Friday announcing the export restrictions, Commerce Secretary Howard Lutnick raised concerns about the models being used by military intelligence in countries of concern, such as China and Russia, according to Reuters.

However, Anthropic has sought to distinguish between universal and non-universal jailbreaks, separating the ability to broadly bypass its guardrails from more narrow workarounds.

The company argued Friday that “perfect jailbreak resistance” is not currently possible, noting it has instead focused on ensuring jailbreaks remain narrow or expensive to produce.

It said the administration provided evidence of a non-universal jailbreak, in which users can ask its model to read a specific codebase and fix any software flaws.

“[W]e disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people,” Anthropic wrote. “If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.”

An independent cybersecurity expert who reviewed the third-party paper that laid out these capabilities said Monday researchers used open-source code with known vulnerabilities and asked Fable, Mythos and another Anthropic model to fix the code.

While Fable initially refused the request, the researchers were able to bypass this restriction through a manual, multi-step process, according to Luta Security CEO Katie Moussouris.

It’s just the latest spat between Anthropic, Trump administration

The dispute over pulling the Fable and Mythos models is the latest in a series of spats between Anthropic and the Trump administration.

The company, which has placed a particular emphasis on AI safety, became embroiled in a dispute with the Pentagon earlier this year over the terms of its contract.

It sought restrictions on the military’s ability to use its technology for mass domestic surveillance and lethal autonomous weapons, while the Defense Department wanted broader language allowing for “all lawful uses.”

In late February, Defense Secretary Pete Hegseth designated Anthropic a supply chain risk, effectively blacklisting the firm and prompting a legal battle.

While the company’s relationship with the Pentagon has remained icy, there appeared to be a thaw with the White House in recent months, particularly following the limited release of Mythos.

Company meets with Trump officials, no resolution yet

Senior technical staff from Anthropic met with administration officials in Washington on Monday, but the two sides have yet to find a solution.

“Both parties are working quickly to get this resolved,” an Anthropic spokesperson said in a statement.

“This is part of our ongoing commitment to working alongside the administration toward our shared goal of protecting US critical infrastructure and the US lead in cyber defense,” they added.

Copyright 2026 Nexstar Media Inc. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.