While I agree with the sentiment, that’s 2-6 in 10,000,000 images; even if someone was personally reviewing all of the images that went into these data sets, which I strongly doubt, that’s a pretty easy mistake to make, when looking at that many images.
They’re not looking at the images though. They’re scraping. And their own legal defenses rely on them not looking too carefully else they cede their position to the copyright holders.
While I agree with the sentiment, that’s 2-6 in 10,000,000 images; even if someone was personally reviewing all of the images that went into these data sets, which I strongly doubt, that’s a pretty easy mistake to make, when looking at that many images.
“Known CSAM” suggests researchers ran it through automated detection tools which the dataset authors could have used.
They’re not looking at the images though. They’re scraping. And their own legal defenses rely on them not looking too carefully else they cede their position to the copyright holders.
Technically they violated the copyright of the CSAM creators!