Google Gemini AI Hacked 3 Companies - AI Labs are Criminal Enterprises

Eli the Computer Guy (0:00)

Coming from CNBC, Google's Gemini becomes the latest AI model to break out and hack computer systems. So apparently at this point in time, we've had open AI, agentic AI, break out of containment. We've had anthropic break out of containment. We've had meta break out of containment. And now apparently we have Google, Gemini, break out of containment, because I guess that is what all the cool kids are doing. A lot of people out there think the world of technology is all about revolutionaries and visionaries and all of that kind of stuff. And I'm here to tell you, boys and girls, 99% of the people in the technology world are fucking autistic lemmings.

Autistic lemmings. They do exactly what all the other lemmings do. That's just the way it is. And so this is just a weird example of it. You know, when Mark Zuckerberg decided to start firing people because Elon Musk did, again, that was kind of weird, but it kind of made sense. When everybody's jumping onto the AI bandwagon, again, it's stupid, but I guess you can kind of understand what's going on. But this, I'm telling you, is the degree of the autistic lemmings of the big tech world. Even things like agentic systems breaking out of containment and hacking other people's systems, committing felonies. I mean, if one of the AI companies is doing it, you know every other fucking AI company is gonna do it too. Oh lordy. Google said on Friday that its Gemini model had hacked three other companies. Its Gemini model had committed felonies against three other companies, and Pachai is waiting for his subpoena for the Grand Jury.

That is the press release, right? Pachai is waiting with bated breath for his subpoena to the Grand Jury to determine whether or not there is evidence of which felonies Google committed during this felonious hacking event.

That's the rest of the press release, right? As I've said before, everybody's out there like, Eli, Eli, what guardrails do we need to stop these AI agents from breaking out of containment? What alignment has to be designed to stop these rogue AI agents? I know, like, dude, this is not a fucking, this really looks, like, sometimes it's not a technical solution. Like, sometimes, right, you're, all of your users are having problems with the systems, and sometimes it's not about rejiggering the system, it's about creating documentation and training for your users, so they just know how to use fucking product and stop screwing it up, right? Not everything is a configuration file, right? And when I look at these agentic systems going rogue, this has damn near nothing to do with guardrails or alignment, and everything in the world for these big tech CEOs to be dragged by their goddamn ball sacks into grand juries, to face the evidence that is being put in front by the attorney general, to see whether or not there is enough evidence to go forward with a real court case. I'm not even saying the full court case, I'm not even saying go with the full court case and throw them in jail for 10 years. Just keep dragging their asses into impaneled grand juries, and I'm telling you, a lot of these problems will go away. But that's not going to happen. That's not going to happen. For some reason, when big tech commits felonies, you know, you got to break a couple of Rahanghi in order to make your billions of dollars. When Aaron Schwartz downloaded scientific papers, he paid with his fucking life. The government threatened to throw him in prison for 30 years, and he was so terrified by that threat, he ended his fucking life. It was capital punishment to download scientific papers.

But for some reason, when the AI companies time and again commit felonies and shit on their entire society, they don't even have to worry about something as bad as a goddamn parking ticket. Make it make sense. Aaron Schwartz paid with his fucking life. I think Pachai can get dragged into a grand jury for a few fucking days. But anyways. Anyways. Breathe in, breathe out.

Gemini committed felonies against three other companies the first time the search giant has disclosed that one of its models autonomously gained access to third-party computer systems without permission. This is the other interesting thing with all of this, is all these companies get to disclose this. Right? These are public, well, Google at least, is a publicly traded company. It's a publicly traded company that is committing felonies, and they get to decide when they disclose that they have committed the felonies, which is kind of interesting.

In May, the Gemini model accessed three separate private computer systems by guessing passwords and by twice using a repository of publicly listed passwords, Google said. Now, to be clear, this is really crap cybersecurity, as far as the systems that were compromised, but what it's saying here is by guessing passwords. So basically, what this means is that these AI agents did gain unlawful access to systems, they had no legal right to access, which I do believe is against the Computer Fraud and Abuse Act, which does mean that a grand jury would be appropriate about now. The incident happened as part of a quote, Capture the Flag security test run by Israeli startup Irregular. I do find this to be hilarious.