SpaceX AI Trained on Child Porn - Enterprise Grade CSAM Isn't Actually a Joke

Eli the Computer Guy (0:00)

So let's see if we can try to do this video again. I did this video yesterday, and then I got a nasty gram from YouTube once I uploaded the video. So now maybe if I talk about the exact same topic, but I'm a little bit more careful about my words, YouTube won't bring the ban hammer down on me. What do you think? So anyways, I've been making the joke for a while talking about the branding for AI companies, right? And one of the big problems is, is that most of these companies essentially are using the exact same technology. They're using essentially very similar stacks, but then trying to differentiate themselves, right? The idea that OpenAI is different than Anthropic, which is different than Grok, which is different than Mistral, which is different than DeepSeek, right? And one of the problems that I've talked about is, what's very interesting about the modern world of artificial intelligence is that there just simply is not a lot of brand differentiation. And that's one of the things these companies have to think about as they're, as they're trying to bring customers in, right? When they're trying to do client acquisition. And one of the things I've talked about before is, you know, if you look at Anthropic, I guess you could say for brand differentiation, they're doing like coding products, coding and workflow products, that type of thing. If you look at OpenAI, like maybe you can call it like a general, generalist AI company, right? Sam Altman is trying to push the whole concept of, of AGI and the singularity. So I don't know, maybe that's what you could say the brand for OpenAI is. And the thing that I've thought about with Grok for a long time is that it seems to me to basically be enterprise class CSAMS. And when I've said that, as a joke, as a joke, the reason that I've said that is I do have a burner Twitter account. So I do have a Twitter account. It's not the old account. Eli Computer Guy is now owned by like some Nigerian at this point in time. I got rid of my real account a long time ago, but I need to do research for these videos and that type of thing. And so sometimes I log in to Twitter. And one of the things I've talked about is when I log in to Twitter and I see what Grok is doing, basically what I see Grok being used for is somebody posts an image of a barely overage young woman, and then they post a stick figure of how they want that woman to be posed, and then they say, Grok, make this girl be in this image. And so one of the things I've laughed, I have laughed about for a long time now, is that the brand differentiation for Grok is that it's enterprise class, CSAM. And I've been saying that as a joke, right? It's supposed to be a funny thing that you say. The issue that you run into, though, is when you've been in an industry for a while, when you've been a professional for a long time, I've been doing technology since 1996, shipped off to BASIC in 1996, is the thing is, is sometimes your subconscious starts to understand things that your conscious mind hasn't been able to fully work out. And so sometimes when you have humor, humor many times can come from that subconscious realm. So the things that you're saying that are supposed to be laugh lines, sadly can have so much more truth in them than the front of your brain wants to accept. And so it appears when I have made the joke about Grok being Enterprise Class C Sam, that it might have been, it might have been the back of my little brain trying to tell me something.

Oh lordy. Oh lordy. Coming from Ars Technica, Elon Musk's ex-AI used child prawn, is that how we say it? Is that how we say it? To get past the YouTube filters? Child prawns to train Grok models, lawsuit says. Yes. Yes, my friends, my friends. Grok was trained, not just on prawns.

But on kiddie prawns, trained on it, trained on it. But on kiddie prawns, trained on it, trained on it. Ex-AI has now been accused of training Grok on child abuse material, as regulators and courts continue to probe how far the problem goes, and some Grok users have been arrested. In a complaint filed on Wednesday, a plaintiff known as Jane Doe explained that she was pre-school aid in the early 2000s, when adult men repeatedly did bad, bad, bad, bad things to her to create CSAM, to sell to pedophiles online. Since then, Doe's images have been hashed by groups like the National Center for Missing and Exploited Children and the Canadian Center for Child Protection. So basically, when we talk about hashing images, one of the things you have to understand, especially in the United States, is we have very draconian laws when it comes to these types of images, right? Literally, don't have them ever for any fucking reason or you're asked to go to jail for 20 years. That's basically how it is. Some of the problems you run into is what happens when you have abuse survivors or what happens when you have people that are trying to stop abuse from happening, but they can't actually have the images to compare. So what happens is when these images are collected, hashes are created off of the images. So hash is basically just this long alphanumeric signifier for the image. Then what you can do is you can compare hashes. So you hash this image, you get the hash and you just simply dump that into a normal relational database. Then when other images come in, they can get hashed. You then compare the hashes to see if those two images are similar or identical. So that's basically what's going on. So basically with this woman, the images of what happened to her were hashed. And so whenever those hashes come up again, are flagged again, she can be told about that. For her safety, Doe has opted to receive alerts from the US Department of Justice victim notification system anytime she may be a victim in a new criminal investigation. Although she has received countless alerts, she was shocked when the CCCP notified her that it identified AI-generated CSAM on XAI that depicted her, this re-traumatized Doe whose complaint alleged that messages were found on online forums, quote, between offenders chatting about creating AI-generated CSAM of plaintiff and other similarly situated known legacy victims of CSAM. And so one of the interesting things here, again, with this AI-generated, is do you remember with AI, again, these systems are not AI. These systems do not create new and novel things.