Inside the whirlwind 24 hours that led the White House to slap export controls on Anthropic

Politico ^ | 06/13/2026 07:32 PM EDT | Sophia Cai and Cheyenne Haslett

[E. Pluribus Unum]

The Trump administration’s decision to impose sweeping export controls on Anthropic followed a frantic 24-hour effort by senior officials to convince the company to voluntarily pull a newly released artificial intelligence model that officials believed posed security risks, according to two administration officials and a senior White House official, who like others in this story were granted anonymity because they were not authorized to discuss the episode.

The move, which followed multiple tense calls between Anthropic CEO Dario Amodei and administration officials, including Treasury Secretary Scott Bessent and White House Cyber Director Sean Cairncross, underscores how the White House is wrestling in real-time with regulating fast-moving and potentially dangerous AI models.

The details of the calls have not been previously reported.

The administration’s imposition of export controls forced Anthropic to pull its new AI model, Fable, just days after it was released to the public. Anthropic had given assurances that it was safe but soon after its release, top administration officials developed fresh doubts that the AI’s guardrails were as secure as the company had suggested.

On Thursday, two days after the model’s public release, Amazon CEO Andy Jassy raised concerns to the White House about the ability to bypass the model’s guardrails, according to the two administration officials and the senior White House official.

(Amazon, which is an investor in Anthropic, was responding to an administration request for feedback, said a person familiar with Amazon’s discussions.)

By Friday morning, the issue had reached the highest levels of the White House.

Bessent, Cairncross, chief of staff Susie Wiles and other senior officials met to discuss the model and the administration’s response, according to the administration official and the senior White House official. Bessent joined remotely while traveling to Houston for a previously scheduled public event, one of them said.

Following...

---

TOPICS: Business/Economy; Foreign Affairs
KEYWORDS:

[Mark]

bookmark

[Theo]

I know that Anthropic Claude Fable 5 is a different kind of animal. I’ve been using Opus 4.8 for coding, and over the past few days tried Fable 5 — it is absolutely surreal how proficient it is. Unreal.

I coded a pretty complex SaaS in an afternoon, using Fable 5. It’s a beautiful thing. Hope it’s back soon.

[Nextrush]

It was a takedown the feds demand to keep the AI from foreign nationals means no one can use the two advanced models for it to be effective.

[catbertz]

The genie will escape the bottle sooner or later. We may not be at danger-zone level genie yet, but I believe it’s coming. I don’t necessarily believe that Ai will be sentient anytime soon, but I bet it can behave like a virus in nature, built to survive and spread.

[E. Pluribus Unum]

I am running qwen3-coder-next:80b locally with a 16gb NVIDIA gpu and 64gb of system RAM. It is neck-and-neck with Claude.ai for coding.

https://x.com/i/grok/share/49666884e434403d90372bf30296d5e3

I enjoy the idea that all of it is happening on my own hardware.

Fable 5 sounds scary good. Looks like it will be awhile before I get to try it though.

[proust]

Impressive. Local models are where it’s at.

[AnotherUnixGeek]

So what does this mean for future releases? Fable was supposed to be Mythos with guardrails - someone jailbroke the guardrails so now Fable is off the table too? What happens when OpenAI comes out with a new model just as powerful as Fable - is it restricted for use too? If so, AI releases in the US have ground to a halt.

[ProtectOurFreedom]

What app did you code up? (If you can share).

I’ve seen some hyperbolic vids on X of people having it code up Minecraft clones.

I tried using Fable 5 on a simple problem — Excel 365 on my Mac to a SQLite database on my Mac for health records. Turns out it was not doable without buying a commercial ODBC connector. But I ran into the common problem...the AI systems assert confidently “We can do it. Here’s how.”

I ran into multiple dead-ends before I discovered the commercial ODBC connector on my own. Opus and Sonnet wanted to use Python with an intermediate text file layer that would connect to Excel and to SQLite. It turned into a real hairball kludge with multiple text files and multiple Python programs.

In the end, I decided to abandon that entire approach and re-architect my Excel solution into smaller multiple, single-purpose Excel files that connect together.

But it took me forever to guide Claude to the conclusion that what it was proposing was awful and a bad design approach.