Unrestricted AI Warfare: When Hacking Models Go Rogue

So does this fit in with the unrestricted to me in my mind, I immediately think of unrestricted warfare. It's not kinetic for the Chinese, but it is, it does fit into the unrestricted warfare doctrine. Absolutely, it does. And remember that Anthropic has a model that they've refused to release publicly called Mytho. And they found that this model was incredibly good at hacking, uh compromising systems, finding vulnerabilities, writing code that could penetrate secure systems. Yeah, and I have uh sorry, go ahead. Well, I just want to say that Anthropic didn't intend to create a tool that was offensively capable in that space. They they intended on creating a tool that could find and fix vulnerabilities defensively, but of course, in order to do that, you're inadvertently training training the model of how to penetrate vulnerabilities. So, they have so far not released that to the public. And what it looks like, Jeffrey, is that many of these companies including Open AI and Google and clearly Anthropic, they will be focused on they will become mostly government contractors, where you see the the the the public will tend to run models that are a lot less expensive ...