The White House and Anthropic: Tensions Over Fable AI

The White House and Anthropic: Tensions Over Fable AI

On Friday night, the White House imposed an export control restriction on Anthropic's Mythos 5 and Fable 5 models, preventing foreign governments and nationals from using these two products. This essentially forced Anthropic to completely cut access to these models, plunging its user base into chaos and casting doubt on the company's future as well as the global prospects for cutting-edge AI.

Over the weekend, several conflicting narratives emerged, although a few points remained constant: allies of the White House claim that just days after the launch, several tech leaders, including Amazon's President and CEO, Andy Jassy, expressed concerns that Fable and Mythos could be circumvented, posing an imminent threat to national cybersecurity.

Both parties discussed the matter on Friday, but the nature and duration of their exchanges vary according to sources: the Washington Post reports that Anthropic had 90 minutes to withdraw its models, while a White House official told Politico that they pleaded with Anthropic "for hours."

Details become even murkier from the White House's perspective. One source told the New York Times that Amazon had found a way to bypass the security safeguards preventing users from employing Fable for cyberattacks. A second source countered to the Times that it was possible to achieve the same results with ChatGPT 5.5 from OpenAI. Semafor reported that it could involve a group linked to China accessing Mythos, although no circumvention has been confirmed. Sources from Axios indicated that the administration simply did not appreciate Anthropic's attitude.

A more in-depth report by Hayden Field, a senior AI journalist for The Verge, presents the viewpoint of Anthropic and the AI security community, which argues that fears regarding circumventions are exaggerated. However, in Washington, the general consensus is that, even if they are ultimately correct, Anthropic and Dario Amodei have created their own political nightmare by opposing the Trump administration — and not in a morally superior way.

One of the most interesting themes emerging is how hard logic and irrefutable facts of science and technology clash with the venality based on political vibes. In the absence of any federal regulatory law, AI regulation entirely depends on the statements that someone has managed to convince Donald Trump to publish on Truth Social or the executive orders he has been persuaded to sign.

As former White House AI advisor Dean Ball pointed out, this reality is what cutting-edge AI companies must operate under. He writes in a scathing post:

• What the law says doesn’t matter.
• What administration officials support one day doesn’t matter.
• Anthropic is a political enemy of this administration, partly because they have explicitly chosen to become one.

And it’s not just a matter of Anthropic and political rivalries with the Trump administration. Anyone at the forefront must understand that in practice, you need an explicit green light from the government.

This mantra certainly applies to all companies operating under the Trump administration, especially those working on emerging technologies. But for some reason, while its competitors have quickly adapted, Anthropic's vibes seem more visibly out of sync with those of the administration — and, tellingly, there are no factions within the administration rushing to defend Anthropic. This inevitably raises questions.