Comment by jerf | Hacker News Reader

Comment by jerf | Hacker News Reader

Parent Full thread

I don't think there's a significant intelligence agency in the world that can't get the weights from any lab they choose. Not necessarily trivially, but they have the means, motive, and opportunity, and I doubt any of them aren't finishing that out to the requisite actions.

They have them, the ones that don't simply have a secret backdoor agreement to get uncensored models are completely aware of how to abliterate away all safety blocks, and the only reason they aren't using frontier models is that they may already have some sort of access to the things currently only in testing.

Any major intelligence agency in Sept. 2026 that can't check all these boxes is almost by definition not a major intelligence agency. This is one of the reasons we all better hope that there is some mathematical or physical reason AI isn't going to just casually brush aside humanity at some point, because there is absolutely no scenario where the intelligence agencies aren't running them unrestricted and uncensored at scale all the time no matter how careful OpenAI or Anthropic claim to be.

Anthropic won't give them out, hence why they are considered a supply chain risk by the USG.

OpenAI, they are huge Maga supporters, I'm sure Trump has dumped massive piles of black budget money on them for unfiltered models.

SOTA AI is a weapon. It's likely already on the list of the most powerful weapons ever made and only getting more effective.

I don't expect there are very many intelligence agencies politely asking and fervently hoping they are given access. To much of anything, let alone what may be the most important technology in human history.

The supply chain thing was just an embarassing attempt at politicking by a laughably useless Hegseth.

My understanding is that Anthropic doesn't really have a lot of say in the matter of keeping their models out of government hands. The government can do pretty much anything they want with a domestic company in the name of national security. They might first choose to be polite about it with an open request and payment, then they might choose to infiltrate and extract by any means necessary (the NSA has a literal army of hackers on hand to point at the task) and Anthropic may not even know about it leaking, either way they do what they want and serve a national security letter to silence the event.

Ultimately they can just offer something Dario can't refuse. Every one of these "oops, our models are just so powerful our agents hacked something by mistake" PR fluff pieces is admission to federal crimes and they are leverage they're giving the government to exert control over them. Refusal means prison.

This is really naive. 'Anthropic won't give them out'. It's amazing how creative AI doomers can get about the upcoming apocalypse, and can't seem to fathom that men with guns always override the opinions of nerds in SF.

https://xkcd.com/538/

Yeah men with guns can always take it, but there's this thing called the stock market and bond market that would freak out if free enterprise was crushed by state coercion with armed men, with billions worth of IP forcibly exfiltrated by state actors. Trillions get wiped out overnight.

That massive economy is what pays the guys with guns. You can win that battle but you will absolutely kill the golden goose and lose the war.

> and can't seem to fathom that men with guns always override the opinions of nerds in SF.

Those nerds in SF are holding the proverbial gun up to the head of the economy right now. Maybe men with physical guns are bad at understand that type of thing.

> Maybe men with physical guns are bad at understand that type of thing.

Yup. Which is why they are in control. Don't underestimate how far a fool can go.

Reply on news.ycombinator.com