Saagar Enjeti

And then it's all on their own to determine how secure the model needs to be to keep other actors out.

And we're really playing with fire here at this point.

Something.

Because it's insane.

I mean, so you've got Anthropic, which is run by, you know, the guy who positions himself as taking safety the most seriously.

And I think with some, you know, at this point he has some credibility given that he refused to go along with some of the Pentagon's demands.

They routinely publish papers that...

do research not only into their own models, but other models to say, okay, this is what we were able to do.

So I do think that they have some credibility there.

At the end of the day, though, this is also a company that is self-interested and that wants to be the first and the best, et cetera.

But we're kind of lucky that it was Anthropic, I guess, that ended up with this extremely powerful AI that

that they took these precautions and realized the problems here.

I mean, that's the other possibility is that you may have a limited amount of testing because there's such a rush to push out the next model and win the race and be ahead of the competitors.

You could imagine a scenario where the testing is not as extensive and they don't even recognize the level of vulnerabilities that they're able to exploit with these models.

But we're putting all of our faith in the hands of like these handful of, I mean, it's just, it's completely, completely insane.

And to your point about, you know, the pushback being, oh, they overhype these products and they claim they're so powerful when really this is just a marketing gimmick.

really see how it serves their interest to not release a model because it's too powerful.

That means you can't profit off of it.

That means that you're really limited in what you're able to do with it.

So I think it's hard to see how this really financially benefits them in the short or medium or long term to if they're really, truly just overhyping this.