The Anthropic Shutdown Was Not About a Jailbreak. New Details Show a Political Fight Dressed Up as a Security Order.

Since the Commerce Department's June 12 shutdown letter arrived at Anthropic at 5:21 PM ET, giving the company a 90-minute window to pull its two flagship models offline, the focus has shifted from what the government said to whether its stated reason holds up.

As of June 16, it largely does not.

What the Government Actually Said

The letter from Commerce Secretary Howard Lutnick directed Anthropic CEO Dario Amodei to suspend all access to Claude Fable 5 and Claude Mythos 5 by any foreign national, anywhere in the world, including Anthropic's own foreign national employees, according to Forbes. Anthropic says the letter provided no specific details of its national security concern.

Anthropic's own public statement, posted June 12, said the company's understanding is that the government believes it found a method to "jailbreak" Fable 5's safeguards to identify software vulnerabilities. The company reviewed the specific demonstration and described what it saw: a small number of previously known, minor vulnerabilities discoverable by other publicly available models without any bypass at all.

The Security Community's Verdict

Katie Moussouris, founder of Luta Security and a veteran cybersecurity researcher, wrote in a blog post that Anthropic shared a private copy of the security paper with her before the shutdown. According to TechCrunch, the paper's authors are security researchers at Amazon.

Moussouris's conclusion: the described bypass should never have triggered an export control. The technical difference amounts to asking a model to review code for security issues versus asking it to fix the code. Functionally similar outputs. She called the order "hasty, heavy-handed, and misguided" and said any attempt to patch the described behavior would only weaken the model for defensive use.

Anthropic's own statement is worth reading directly: "We suspect that perfect jailbreak resistance is not currently possible for any model provider. Every safeguard used in the industry is vulnerable to non-universal jailbreaks." The company stated this openly at Fable 5's launch. It wasn't hiding the limitation.

What Was Actually Going On

The technical case for the shutdown was thin from the start. Axios, cited by TechCrunch, reported something more revealing: sources described the directive as driven largely by "personality differences" between Anthropic and the Trump administration, not by the jailbreak itself.

The Verge, reporting on the weekend negotiations between Anthropic and the White House, adds context. This fight did not start on June 12. Anthropic earlier this year refused to allow U.S. military use of its models for domestic surveillance and fully autonomous weapons systems. The government put Anthropic on a supply chain blacklist, according to The Guardian, set to take effect later this year.

Friday's export control order landed as that broader dispute had reportedly been showing signs of easing. The timing suggests the jailbreak report gave the administration a legal mechanism that the underlying political conflict may have been looking for.

White House adviser David Sacks claimed Anthropic refused to fix the issue. Anthropic's position is that the fix Sacks described isn't technically meaningful and would degrade the model's legitimate defensive capabilities.

The Strongest Case for the Government's Action

The administration's concern is not invented from nothing. Mythos 5 was restricted to a select group of government agencies and companies precisely because Anthropic itself labeled the underlying model too dangerous for general release. When Anthropic publicly warns that a model's predecessor was too risky to deploy, it invites exactly this kind of scrutiny.

If the full Mythos 5 architecture sits underneath Fable 5, and if a jailbreak can peel away Fable's additional safeguards and expose Mythos-level capabilities, that is a legitimate national security question. The government also had Amazon's security researchers raising the alarm. A cautious reading of the facts could justify demanding answers before letting hundreds of millions of users, including foreign nationals, continue accessing the model.

The problem is the process, not the underlying concern. A 90-minute ultimatum, a letter without specific technical detail, and a unilateral shutdown with no apparent court involvement or formal finding is not what careful national security regulation looks like.

The Precedent

TechCrunch framed this plainly: the U.S. government successfully forced a tech company to pull its products offline with a swift, unilateral action that did not appear to require court approval. That mechanism exists independent of whether Anthropic deserved it.

Every other AI lab is watching. OpenAI, Google DeepMind, and Meta all deploy models to hundreds of millions of users globally, including foreign nationals. If a single Commerce Department letter can shut that down overnight, based on an unspecified finding that the company being shut down disputes, the entire commercial AI industry now knows it operates at the administration's discretion.

The Pentagon's chief information officer, Kirsten Davies, posted in support of the directive. Dozens of security researchers, per TechCrunch, have signed on to a public call demanding the Trump administration reverse the order.

Anthropic's executives met with Trump administration officials over the weekend, per our prior coverage. As of June 16, no resolution has been publicly announced, no charges or formal legal proceedings have been filed, and the models remain offline. The specific question left hanging: whether the administration will provide Anthropic the technical specifics it has been demanding, or whether the shutdown continues indefinitely on the basis of a classified finding the company is not allowed to see or rebut.