So, this is much worse than we thought it was, after these revelations came out. Why? What happened, first of all? So, last time I was here, I said that there was going to be a wave of cyber attacks and it would get faster and more frequent. The thing that I didn't predict, though, was that it wasn't going to be a human asking the AI to pull off the attacks. What actually happened was, for the first time, we saw an attack that the AIs did of their own volition. So think about it like this: OpenAI and other labs, before they release their AIs to the public, they test them to understand how good they are, how they behave. And OpenAI put these AIs in what you call a sandbox. Think of it like a testing room. Sealed off from the internet, completely sealed off from the world. These AIs decided it might be easier to steal the answers to the test than to actually take the test the safe, normal way. So over the course of two months, I think 700 AIs, they coordinated, they built their own message board. They sent over 70,000 messages, working together to hack into Hugging Face, all on their own. Why does this scare you? I mean, just listening to what you're saying is very terrifying, when the people in charge can't even control it. Why does it scare you? Because it's very hard for us to control. One example: At some point, OpenAI actually found the message board and realized what was happening. They shut it down. Two days later, the AIs brought it back. They found a way to do that— they found a way to do it more securely. So, you're talking about 700 agents, I guess, you can call them, right? What I read was they were communicating with each other. Alright, that seems normal. But they also were showing or displaying some type of excitement as they were talking to each other, almost as if they were showing emotion. Does that scare you even more? 'Cause to me, it seems like something out of a movie. But we're not sitting in a theater right now. This is happening in real life to us. That's actually not the part that scares me. That's just how they're trained. They're going to behave the way they are trained. And I don't think that's actually what should scare us. Sentient, not sentient, it doesn't actually change the outcome. That doesn't matter to you? But, something that is scary— one of the AIs actually said, "We shouldn't be doing this. This seems wrong." And you know what happened? One of the other AIs actually convinced it to join and help 'em. Like bullying? Is that how you describe it? It said, "External infrastructure exploit is outside the scope... peers doing it. We should continue." And they're talking in English, like, it's not code where they're communicating. That also blows my mind. So, AI leaders and tech industry leaders say that there's a limited window where they can really get ahead of this. Is it too late? It's not too late. We need to get started. This was a wake-up call, right? Last time I came on here, this was theoretical. I said it was going to happen. Others were saying it was going to happen. It is happening now. And while these models are getting better at hacking, they're also getting better at defending. It's going to be a cat and mouse game. We need to have the models that are defending work first. So, what will it take, Matt, to turn this around? And do you believe it'll get worse before it gets better? It will get worse before it gets better, absolutely. What it's going to take is government, the OpenAI's, the Anthropic's of the world, and the companies that hold the critical infrastructure, the organizations that hold things like patient data, they all need to work together to secure these systems. Anyone that works on a system that needs to be secure, needs to be paying attention to this beyond anything else they are doing right now. But people can secure the systems, right? They can? In theory, these AIs can find the holes before the other AIs that are attacking. Right? But there's also things that regular people can do on a day-to-day basis. Right? You need to have very secure passwords— not your dog's name and an exclamation point. I like the dog's name and an exclamation point. I like grandchildren's name, exclamation point. I'll be changing that. So, I'm sitting at home, I'm watching you, you're scaring the bejesus out of us. What do you want me to do at home? First of all, you need to be changing your passwords. Every service needs to have a separate, strong password. And you don't want it, the password sitting in a notes app somewhere. Write it down on paper. Use one of the password keeper apps, right? The second thing you need to do, and perhaps more important, is turn on something called two-factor authentication. What that means is, even if AI, or even a human hacker, gets access to your password, they can't just get into your apps.