Trump administration imposes export controls on Anthropic following a dispute over bypassable guardrails in its Fable 5 model

https://www.politico.com/news/2026/06/13/inside-the-whirlwind-24-hours-that-led-the-white-house-to-slap-export-controls-on-anthropic-00961519?gpp=DBABLA~BVQqAAAAAgA.QA

@cheyennehaslett You can only think this is “unprecedented” if you discount export controls on hundreds of other novel and dual-use technologies. What a dumb and sloppy characterization.

@kimmonismus You can't have a super weapon and then refuse to work with the Pentagon, but then also sell your weapon worldwide.

Imagine the CEO of Raytheon trying that.

The sheer gall to tell the country that made you, to f*ck off, was reckless arrogance.

@kimmonismus Anthropic and it's holier-than-thou attitude is just irritating. Most people like this end up doing the opposite of what they preach.

@Solaawodiya i get your point. we still try to figure out which side is true

It's clear that Anthropic expected one-off jailbreaks to occur -- because it's impractical to stop them. Instead, they focused on stopping a universal jailbreak.

It's also clear that the Admin is freaked out by the one-off jailbreaks - perhaps equating them to a "bug" that can be simply patched.

This leaves the state of the AI frontier at an impasse. The Admin is demanding a level of "safety" that Labs simply cannot deliver.

The result is that progress at the frontier will be stifled - expanding the frontier invites political headaches without economic benefits.

@cheyennehaslett Sounds like openAI has always had better guardrails and alignment team than anthropic, who mainly talk about it. They had months time to fine tune fable and then released it 2 weeks after opus 4.8 blocking all medicine bio chemistry Qs. Why released so fast? IPO soon?

@cheyennehaslett the problem is none of those people talking with him are technical. the cyber director? he's a lawyer. Who on that call was technical enough to have an intelligent conversation with Dario? Could they even explain the point they were making?

@kimmonismus Have you noticed the tendency of the White House to lie about just about everything?

That might help tip the balance of who's more likely to be telling the truth, or close to the truth.

@kimmonismus Katie has seen the paper.

@kimmonismus Tech execs love apocalyptic rhetoric until they’re asked to act

Much here is about the model finding exploits in software which have high impact, like stealing $100m on a blockchain, attacking widely used infrastructure, etc. The power to find those issues can be used to fix or to exploit them, but how should the rail guard know. I can imagine the guards can be better trained to make such decisions but this seems kind of a new area.

@cheyennehaslett It’ll be interesting to see how markets react as they digest the meaning of Trump’s banning of Fable. The implication for the US AI industry is chilling: the govt may step in to slow the advance of AI tech, upon which the entire AI complex’s valuation multiples depend.

@edwinhayward I cover both sides, as neutral as possible.

@kimmonismus All of you guys have been to quick to judge Anthropic and call them fools…

@AndeanViber @kimmonismus Great argument 👍 you have convinced me. Fucking idiot.

What Anthropic said was "yes, we will cooperate on everything, like the Venezuela operation where Claude played a significant part.+, except mass surveillance or autonomous weapons because that would break the law". - >

@kimmonismus Once again .. it shows that Dario is a retard .. What a dumb ass.

@cheyennehaslett This was clearly a case of Trump exacting revenge on @AnthropicAI for their refusal to bow down to @PeteHegseth in Feb and subsequently fighting the admin in court. This was not about “national security.” It’s also, as are many things Trump, a massive own goal.

@kimmonismus @pangram