Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery • Hacker News Board

Re: Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

#261

Post by FireBeyond » Thu, Aug 06, 2026, 1:34 PM UTC

unsnap_biceps wrote: post #161

ihuk wrote: post #65

Earlier quoted context omitted.

Reported an obvious sexual ad the other day and got a very AI generated response that started with "Dear Username" stating that they reviewed the ad and found nothing wrong with it.

Before I gave up on Facebook entirely, I reported an ad that was an explicit blow job and received a similar response and this was before AI was huge. I really don't think they've ever cared.

My example was a comment that said "only a shame more [hard r n word]s didn't die. we need to get this place cleaned up" and told that was not in violation of community standards.

Re: Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

#262

Post by specproc » Thu, Aug 06, 2026, 1:36 PM UTC

inigyou wrote: post #242

WorldMaker wrote: post #234

Earlier quoted context omitted.

Section 230 should have never applied to ads. Ads are a commercial relationship. Facebook and YouTube gets paid by these scammers so why would they want to accurately police them? The incentives are broken. Newspapers had to invent all sorts of ways to make sure that they had clear markers between ads, editorial content, and sponsored editorial content. Such things were enforced by lawsuits and a couple of Ad Bureaus…

Section 230 doesn't say the advertiser can't go to jail or Facebook doesn't have to tell the police who the advertiser is.

Who goes to jail for displaying the ad in a public place? If I ran say, a shop, and put CSAM up on my notice board (because I was paid to) that would see me in prison, surely.

The lack of accountability here is astounding.

Re: Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

#263

Post by stephenbez » Thu, Aug 06, 2026, 2:17 PM UTC

Spooky23 wrote: post #203

ryanjshaw wrote: post #135

Earlier quoted context omitted.

Google Play Store showed my child an ad of a bloody body on a bed for a murder mystery game. The game was rated 18+ and the account had my daughter’s correct single digit age. I reported the ad and a week later it was still there. I switched to Apple.

Yet you can’t run Linux on a Chromebook with a minor child user.

What do you mean? You can use Linux on a Chromebook just fine if the user is an adult but if the user is a child there is some Chromebook restriction?

Re: Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

#264

Post by nticompass » Thu, Aug 06, 2026, 3:01 PM UTC

craftkiller wrote: post #60

I've noticed on YouTube I see a lot of adult sexual ads. Things like home cleaning ads that have an unmistakable wood-grain image of fellatio. I've always wondered "how are these getting past the moderators?". I guess the answer is the obvious one: no one is moderating anything.

I've noticed this with obvious scam ads on the YouTube app or even Google search results. I've determined that (much like Twitter/X) the "verification" (or "moderation") is simply just "did they [the advertiser] pay us?"

Re: Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

#265

Post by remuskaos » Thu, Aug 06, 2026, 3:06 PM UTC

ryanjshaw wrote: post #135

craftkiller wrote: post #60

I've noticed on YouTube I see a lot of adult sexual ads. Things like home cleaning ads that have an unmistakable wood-grain image of fellatio. I've always wondered "how are these getting past the moderators?". I guess the answer is the obvious one: no one is moderating anything.

Google Play Store showed my child an ad of a bloody body on a bed for a murder mystery game. The game was rated 18+ and the account had my daughter’s correct single digit age. I reported the ad and a week later it was still there. I switched to Apple.

The play store is pretty bad. About half the time I search for an app, BY NAME, without any way to confuse it with another app, the top result is a scam or copy cat, or adware. And always, without fail, they are paid-for ads instead of real search results. How can it still be somehow advantageous to trick your users into installing an app they clearly don't want. I wish there was a real alternative to the play store.

F Droid comes close and can even auto update now (never understood why that didn't work on the past), but most apps I want or need like banking, health insurance etc aren't on F Droid.

Re: Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

#266

Post by FireBeyond » Thu, Aug 06, 2026, 3:12 PM UTC

Ajedi32 wrote: post #216

callc wrote: post #189

Earlier quoted context omitted.

The problem is equating profitability with what societies want and don’t want. Morality exists outside of the business world (and it should inside as well). Further, it is a societies’ right to enforce the boundaries of what business are allow to do / not do. For example: banning ads along roads and scenic areas. Idk, the whole idea of “how much fines will I get doing X shitty business practice” is just gross. I’d ra…

Perhaps, but when you're operating at any significant scale accidents become a statistical certainty. 0 screw ups, ever is not a realistically achievable goal. And if some limited number of screw ups is allowable, that necessarily becomes an insurable cost that amounts to "just the cost of doing business", so I don't really see the complaint. Concrete example: If you run a construction business, even if you take all…

Yes, screw ups become a certainty. But these are screw ups that the company doesn't meaningfully care to fix. Just as we'd have an issue with a company that has OSHA findings that they continue to ignore. But with Facebook, "Our automated review found nothing wrong with [this ad that implies a young child performing sexual activities]".

Also, speaking of statistics, what?

> statistically you're going to get 9.2 of your employees killed for every 100,000 man-hours they work.

No, you're not. This would mean if you employed 50 full-time employees you'd expect 9 of them to die per year.

What that graph actually says: "In a population where there are 100,000 construction workers, you can expect 9 fatal workplace accidents per year".

Re: Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

#267

Post by mullingitover » Thu, Aug 06, 2026, 3:24 PM UTC

WorldMaker wrote: post #234

Spooky23 wrote: post #209

Earlier quoted context omitted.

Absolutely. My uncle got taken by an “investment club” supposedly endorsed by Warren Buffet and run day to day by some real financial person. All communications were via WhatsApp. They started with real stock tips (which were pretty good actually!) and then did a pump and dump. See: https://www.cleuremissionfund.com/ I received a similar ad (mine was from the shark tank people), reported it and got a message similar…

Section 230 should have never applied to ads. Ads are a commercial relationship. Facebook and YouTube gets paid by these scammers so why would they want to accurately police them? The incentives are broken. Newspapers had to invent all sorts of ways to make sure that they had clear markers between ads, editorial content, and sponsored editorial content. Such things were enforced by lawsuits and a couple of Ad Bureaus…

In my case I wasn't even reporting ads. These are plain old Facebook accounts and groups using very obvious AI-generated slop images of pets for 'adoption' or 'rehoming', so Facebook isn't even making money on this directly. They are participating in the fraud purely out of the malice (more realistically, incompetence spiced with greed) in their own hearts.

It's been a very well known problem on Facebook for years[1][2], and the pathetic part is that its so obvious. It really tells you how bad Facebook is with AI, because even I as a regular 1x engineer could wire up some very cheap inference that would lay waste to this fraud pattern as a part-time weekend project.

[1] https://consumer.ftc.gov/consumer-alerts/2024/05/looking-buy...

[2] https://www.bbb.org/scamtracker/lookupscam/807274

Re: Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

#268

Post by cindyllm » Thu, Aug 06, 2026, 3:24 PM UTC

inigyou wrote: post #257

gs17 wrote: post #141

Earlier quoted context omitted.

Interesting, I've kind of made a hobby of reporting scams and hate speech and watching it take a week to come back with "it's no violation", requesting another and getting the same response in another week. I've never seen them do anything no matter how obvious or awful it is.

You have to actually, honestly, fall for the scam and lose all your money and only then will you have standing to sue Meta for making you lose all your money. You won't be able to afford to sue them, of course.

[dead]

Re: Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

#269

Post by tremon » Thu, Aug 06, 2026, 3:37 PM UTC

remuskaos wrote: post #265

ryanjshaw wrote: post #135

Earlier quoted context omitted.

Google Play Store showed my child an ad of a bloody body on a bed for a murder mystery game. The game was rated 18+ and the account had my daughter’s correct single digit age. I reported the ad and a week later it was still there. I switched to Apple.

The play store is pretty bad. About half the time I search for an app, BY NAME, without any way to confuse it with another app, the top result is a scam or copy cat, or adware. And always, without fail, they are paid-for ads instead of real search results. How can it still be somehow advantageous to trick your users into installing an app they clearly don't want. I wish there was a real alternative to the play store.…

> never understood why that didn't work on the past

AIUI, Android simply didn't support it. The original way for apps to install other apps was via ACTION_INSTALL_PACKAGE, which is a user-visible action. The InstallPackage API did not support silent updates until sdk version 31: https://developer.android.com/reference/android/content/pm/P...

Re: Meta Ran Ads That Contained AI-Generated Child Sexual Abuse Imagery

#270

Post by Ajedi32 » Thu, Aug 06, 2026, 3:43 PM UTC

FireBeyond wrote: post #266

Ajedi32 wrote: post #216

Earlier quoted context omitted.

Perhaps, but when you're operating at any significant scale accidents become a statistical certainty. 0 screw ups, ever is not a realistically achievable goal. And if some limited number of screw ups is allowable, that necessarily becomes an insurable cost that amounts to "just the cost of doing business", so I don't really see the complaint. Concrete example: If you run a construction business, even if you take all…

Yes, screw ups become a certainty. But these are screw ups that the company doesn't meaningfully care to fix. Just as we'd have an issue with a company that has OSHA findings that they continue to ignore. But with Facebook, "Our automated review found nothing wrong with [this ad that implies a young child performing sexual activities]". Also, speaking of statistics, what? > statistically you're going to get 9.2 of yo…

Yes sorry, man-workyears not man-hours.

> "Our automated review found nothing wrong with [this ad that implies a young child performing sexual activities]"

That's also a screw-up, and they should be fined accordingly, proportional to the amount of harm this screw up is causing to society. If there are any practical steps they can take to fix the problem, that should incentivize them to do so.

But if for whatever reason it turns out the only possible way to fix such screw-ups is for them to hire 100,000 employees to manually review every single ad, and the cost of that both exceeds the cost of paying the fines and the opportunity cost of turning off this feature completely, so they choose not to do that, is that a problem?

It's like if you tried to insist construction companies evacuate all workers from the construction site before operating any heavy machinery. Maybe that would save some lives, but it would also be completely impractical.