I wonder if you can tell us just how deep that relationship goes.

Yeah, so we, we, through a Freedom of Information Act lawsuit, um, got a set of, uh, contracts that, uh, OpenAI, Anthropic, Google, and, um, xAI, uh, Elon Musk's, uh, uh, outfit, signed with the Pentagon last summer to develop militarized, um, uh, LLMs, uh, for, for, um, for the Pentagon's use.

Um, and You know, they were heavily redacted, but what was, what really came through clearly from the unredacted portions was that they, these companies did not just sell the Pentagon a product, but they sold them a very, very intimate relationship, an ongoing collaborative relationship wherein, um, they were advising the Pentagon, Pentagon essentially on how to use these products to wage war.

Um, they were getting classified briefings.

They were getting briefings on, uh, US military, military operations.

Um, they were providing hands-on training.

They were providing threat assessments.

How is that different from the kinds of police support for ICE we've seen in other areas?

I wonder if you can tell us just how deep that relationship goes.

Yeah, so we, we through a Freedom of Information Act lawsuit, um, got a set of, uh, contracts that, uh, OpenAI, Anthropic, Google, and, um, xAI, uh, Elon Musk's, uh, uh, outfit, signed with the Pentagon last summer to develop militarized, um, uh, LLMs, uh, for, for, um, for the Pentagon's use.

Um, and- You know, they were heavily redacted, but what was, what really came through clearly from the unredacted portions was that they, these companies did not just sell the Pentagon a product, but they sold them a very, very intimate relationship, an ongoing collaborative relationship wherein, um, they were advising the Pentagon, Pentagon essentially on how to use these products to wage war.

Um, they were getting classified briefings.

They were getting briefings on, uh, US military, military operations.

Um, they were providing hands-on training.

They were providing threat assessments.

That's how this technology takes hold, and the, the private companies say, "Go try it." The cops become essentially unpaid marketing for these companies, and then suddenly they're in tens of thousands of cities without much oversight.

What sort of things were they looking into, and what kind of content online were they determining was sort of domestic extremism?

So we don't know what they were looking for.

They probably pulled a few examples out of a much broader set of posts that they were monitoring or, or had, you know, saved for later analysis or investigation.

It's a great question.

Like, are they just reading every tweet that mentions data centers? Or, you know, and probably using a tool like Dataminr or some other sort of, you know, sentiment analysis stuff to do, like, a pretty broad search.

I don't know.

But what we- what they flagged in the memo, it's a pretty wide range.

It's not, like, these, like, people that are like, "I've looked at the permitting documents for this, and, and I think we can stop the data center through this sort of, like, niche loophole." Like, those people just, I've never seen them sort of make violent rhetoric.

Sam, what do we know about how the Pentagon is using AI tools in the Trump administration's various wars?

So under, uh, Secretary of Defense Pete Hegseth, there has been a huge, very aggressive push to integrate AI really wherever and whenever possible.

Um, I think that you're sort of seeing the Pentagon under Hegseth mimic a lot of tech industry rhetoric, which is like, we don't totally understand the, this technology.

We don't totally know where it's gonna be useful, but we need to use it as much as possible anyway.

I think that you've seen DoD under Hegseth be extremely aggressive in the cadence of airstrikes.

Um, this is a Pentagon that believes in killing people just sort of, I think at, at times it seems to sort of, sort of give, um, itself things to tweet about.

Um, this is, you know, a political movement, uh, and, and an ideology guiding the Pentagon that I think relishes violence and these AI systems, you know, when you wanna blow things up and kill people, these tools can provide a very rapid turnkey means of having a list of people and places to destroy.

I will also note, in 2024, The Intercept sued OpenAI in federal court over the company's use of copyrighted articles to train its chatbot, ChatGPT, and the case is ongoing.

Right

... has been debunked a million times over.

I do not think that that is the risk presented by Roblox, that it's going to, like, brainwash kids or harm them that overtly.

I think that you really nailed it.

Those things were so obviously, like, fantasy and make believe, whereas Roblox really wants to encroach upon the real world.

Like, it wants to be the place where your actual friendships play out.

When you weren't playing The Sims, you were probably hanging out with your friends in person, right? Maybe you guys would play The Sims together.

And I, I often wonder if I should join the lawsuit.