The US government’s ban on human models was never about jailbreaking AI

An enforcement letter sent by the US government to Anthropic effectively forced the company to do so Pulling its latest AI models offline Just before the weekend, this should be a wake-up call for any American tech company — AI lab or otherwise.

To catch up on the breaking news: On Friday afternoon, the US Department of Commerce sent Anthropic a letter citing a vague export control directive barring non-Americans, including Anthropic employees, from accessing Fable 5 and Mythos 5, citing unspecified national security concerns. Anthropic said it believes the message is related to bypassing the model’s guardrails, but is unsure because the message does not provide specific details. The letter was not made public.

In response, Anthropy close Both models are premium to all customers to ensure they comply with the directive. The result was that the US government succeeded in forcing a technology company to take its models off the Internet with a quick and unilateral action that did not appear to require court approval.

The Trump administration’s intervention on Friday shows that the AI ​​industry is not immune to government interference. It’s also a warning to the broader tech industry: comply, or we can shut you and your products down.

Quoting sources, Axios He described a tense situation over the weekend between major players, saying “personal differences” between Anthropic and the Trump administration led to the export directive, rather than a technical issue with the AI ​​products.

New details about the case that emerged over the weekend now cast further doubt on the government’s already shaky logic.

said Katie Moussouris, a cybersecurity expert and researcher who founded Luta Security In a blog post Anthropic recently shared a proprietary version of a research paper written by security researchers describing an alleged firewall bypass in Fable 5. (The Wall Street Journal reports that the paper’s authors They are security researchers at Amazon.) Moussouris said Anthropic contacted her to ask her to take the paper.

Moussouris’s blog post described how researchers activated the guardrail bypass, but said the bypass itself “should never have triggered export controls.” The difference is largely between asking an AI model to “review code for security issues” versus asking it to “fix this code.” The end result is pretty much the same, even if the questions are asked a little differently.

“The behavior described in the paper cannot be meaningfully reformed, and any attempt will only weaken the defense model,” said Moussouris, who criticized the export control directives as hasty, heavy-handed and misleading.

Moussouris and dozens of other leading researchers and security experts have since called on the Trump administration to do so Cancellation of an export control orderHe called the move to strip advanced cybersecurity capabilities from U.S. network defenders “dangerous.”

Previous administrations have made comprehensive decisions on knowledge gaps. For example, the language used by the US government during the 2000s to reform export law covering cybersecurity tools that could also be used in cyberattacks was so broad that Unintentionally almost banned Legitimate security and vulnerability research.

However, the Trump administration’s directives appear to be retaliatory.

Justin Hendricks, Editor of Tech Policy PressHe said the Trump administration’s move “will likely raise alarms in foreign capitals about the reliability of American artificial intelligence in important applications.” The message is that US AI companies cannot be trusted to operate without interference from the US government.

The Trump administration has not confirmed why it activated the export control directives. Did officials misread the report and panic? Is Amazon CEO Andy Jassy Say something to senior government officials Who prompted the reaction, out of caution or spite? Was there something missing in translation, or was this a way to put pressure on the Anthropists, who the administration was already with? He has a tense relationship? It is possible that the White House was unaware of the far-reaching consequences of the letter request and that officials are scrambling to repair the damage they caused.

As Hendricks puts it, “The climate is clouded with suspicion that senior officials are choosing preferred candidates on the basis of personal and political factors.” The result is that the government has set a dangerous precedent over how much control it intends to exercise over the launch of American-made software.

This time the government objected to Anthropic; Tomorrow could be with anyone else.