Those same images have made it easier for AI systems to produce realistic and explicit imagery of fake children as well as transform social media photos of fully clothed real teens into nudes, much to the alarm of schools and law enforcement around the world.
Until recently, anti-abuse researchers thought the only way that some unchecked AI tools produced abusive imagery of children was by essentially combining what they’ve learned from two separate buckets of online images — adult pornography and benign photos of kids.
But the Stanford Internet Observatory found more than 3,200 images of suspected child sexual abuse in the giant AI database LAION, an index of online images and captions that’s been used to train leading AI image-makers such as Stable Diffusion. The watchdog group based at Stanford University worked with the Canadian Centre for Child Protection and other anti-abuse charities to identify the illegal material and report the original photo links to law enforcement.
Wow, calm the fuck down dude.
The reason they didn’t know is because the AI groups aren’t the ones scanning the Internet, different projects do that and publish the data, and yet a different project identifies images and extracts alt text from them.
They’re probably freaking out about as much as any search engine is when they discover they indexed CSAM, and probably less because they’re not actually holding the images.
I know the point you’re going for, and raging out at the topic only undermines your point.
“Other groups organized this data, but we couldn’t be fucked to check to make sure it was all fully legal and above board” said nobody who actually cared about such things ever.
The fact that they don’t check because it would take too long and slow them down compared to competitors is literally the point. It’s all about profit motive over safety or even basic checking of things beforehand.
It’s a really, really weak excuse.
Did you know that they actually do check? It’s true! There’s a big difference between what happened, which is CSAM was found in the foundation data, and that CSAM then being used for training.
Also, the people who maintain the foundational dataset do checks, although which was mentioned by the people who reported the issue. Their critique was that the checks had flaws, not that they didn’t exist.
So if your only issue is that they didn’t check, well… You’re wrong.
400 million images that is. Checking all is impossible.