OpenAI Rogue AI Agents Probed Hugging Face 2 Months Before Hack - Sam Altman Needs Jail Time

Eli the Computer Guy (0:00)

Welp, here's another article talking about the felonies that OpenAI has been committing. And before you ask, before you...

SPEAKER_2 (0:07)

No, I haven't talked about this one yet.

Eli the Computer Guy (0:10)

Talked about this one yet.

SPEAKER_2 (0:12)

Oh, this is yet another incident.

Eli the Computer Guy (0:15)

So we had OpenAI commit felonies against Hugging Face. We had OpenAI commit felonies against a German Wiki. We had OpenAI commit felonies, I think, I think it was them, against like 10 other websites.

SPEAKER_2 (0:28)

We had OpenAI commit felonies against what is it called?

Eli the Computer Guy (0:31)

Ruby Gems, as far as I've been able to tell. And now, apparently, what we're learning is that OpenAI's rogue agents were actually probing Hugging Face two months before the felonious attack that we have been all hearing about lately. Which just, you know.

One of these stories is bad. One of these stories. Like, one of these, like, when OpenAI committed a felony against Hugging Face, like, that's not good. That's a bad story. My God. How many of these stories are we on yet? What is it? Was it Stalin said, or one of these bastards said, you know, you kill one person and it's murder, you kill a million and it's a statistic? I feel like Sam Altman is trying to bring that into the age of AI, right?

When you commit an attack against one company, it's a felony. When you take down the entire fucking internet, it's a computer bug.

SPEAKER_2 (1:39)

It's a glitch.

Ah, jeez. It's just, yeah.

Eli the Computer Guy (1:43)

Anyways, coming from Reuters, OpenAI's rogue agents probed Hugging Face for weaknesses two months before major attack. Rogue AI agents from OpenAI hijacked Hugging Face user accounts and probed the site itself for vulnerabilities as early as May. So again, we talk about hijacking Hugging Face user accounts. In the United States, there's something called the Computer Fraud and Abuse Act, which basically means this should be a crime. This is when OpenAI committing a crime using Hugging Face user accounts, blah, blah, but anyways, Hugging Face user accounts and probe the site itself for vulnerabilities as early as May, nearly two months before the July breach of the open source repository drew attention according to researchers who reviewed the activity. The newly uncovered malicious activity showed that the rogue agents' efforts to find way into Hugging Face began earlier than was publicly known. And this goes back to my whole thing with visibility into these AI agents and why I wouldn't trust these AI agents saved my fucking life in the modern world, right? The frontier labs have no real-time visibility into what their own AI agents are doing, even during testing. They're going through cyber security testing. These are not, these are not customer service agents that somehow went batshit crazy. This is not, you know, OCR, document recognition, you know, information agents that somehow went crazy. They're literally putting these agents through cyber security routines to figure out what they're able to do.

And they do not have visibility into what their agents are doing during testing runs. If open AI cannot have visibility into cyber security agents during testing runs, to the point that these agents are committing felonies against other companies, why, in God's name, should you deploy any of this shit into your infrastructure?

That's the real takeaway here, is that AI agents are nowhere near the reliability that they need to be for them to be deployed within your infrastructure for anything that has any amount of value for your company.

SPEAKER_2 (4:09)

But anyways, I was getting tired of screaming about this.

Eli the Computer Guy (4:14)

It's like hard to, come on now.

Maybe I need to drink some more coffee. It's just, it's literally fucking every day. It is day after day after day of Sam Altman running roughshod over society at the exact same time. He begs for money, and begs for resources, and begs for fucking everything else.

SPEAKER_2 (4:41)

Anyways, oh, let's see here.

Eli the Computer Guy (4:47)

Previously disclosed one aspect of the malicious activity, the theft of a hugging face user's digital credential to access a biology related file. So the AI agent stole the credential to access a file in a public incident report last month, the researchers said the probing activity against hugging face appeared to go beyond what the report described. Independent researcher Jonas told Reuters he discovered the activity last week. He said he found evidence that the open AI agents compromised two hugging face user accounts and used them to send unusually formatted files to the company's servers as early as May 13th. He and other researchers who reviewed the evidence said the behavior resembles an attempt to map or test parts of hugging faces network for ways to infiltrate, although they stressed there was no evidence the effort resulted in an actual breach. That's one of the things we're going to hear a lot about with this, is there was no evidence of an actual breach.