Wow, I'd guess the opposite when we consider the base rate. Seems like a classic Bayesian problem.

It's fairly normal for parents to take pictures of their children naked in the bathtub/at the beach/ camping/etc.

Conversely, I'd expect actual pedophiles and CSAM producers to be really quite rare.

So even a relatively low base-rate of normal parents with normal nude photos would likely dwarf CSAM upon detection.

So, if we say 1/100 are pedophiles, and 30/100 are parents, and of the parents 10% have such photos, the ratio I'd expect without getting into detection rates is like 3:1 in favor of normal parents.

If it wasn't clear, people aren't objecting to the use of automated tools to prevent crime. It's that there is absolutely no avenue of appeal or review against it even if the law enforcement exonerates them and a big news media reports it.

Google has inserted itself into almost all spheres of digital life by hook or crook. It's practically difficult to avoid them in many services - especially email. And now they play judge, jury and executioner. I don't understand any of these are acceptable, much less justifiable. The old argument 'think of the kids' used to justify digital authoritarianism is such a cliché by now.

I’d put pretty good money on almost every family having pictures of their naked kids doing some shenanigans. I know I have those pictures. My parents have those pictures of me and my sister. Quite a few, as I seemed to enjoy trying to run about naked…

Bill Waterson somehow managed to sneak watercolor paintings of a naked little boy into every major newspaper under the guise of being a “comic strip”— the perv.

Somehow I think there more parents who sometimes need to take a photo of their naked children than there are paedophiles.

Or at least the ratio is clearly not 1:100000, maybe more like closer to 1:10.

You would need statistics how many times google have reported police and how many times it have turned out to be a false alarm. Does google even keep record of false alarms? Most likely they don't to avoid responsibility.

Well as a data point I have pictures of my children naked. As another data point my parents have photos of me as a child naked, and as a third data point my grandparents have photos of themselves as children naked. Whereas I don't knowingly know any paedophiles.