Why doesn’t Facebook just ban AI slop like Shrimp Jesus?

These tend to be posted on Facebook pages with generic-sounding names and captions that don’t necessarily reflect the photo, like “why don’t photos like this ever trend,” or “beautiful cabin crew,” and “Scarlett Johansson” for an image of Jesus built into the hood of a car.

Why slop? The best explanation is that the Facebook pages posting these are making bait for potential scam targets, helping scammers identify gullible people in the comments who they can potentially extract money from.

Another possibility is that posting these images is a good tactic to build up a large page audience, which can be lucrative.

The vague similarities among the pages suggest that the AI slop is being posted by a network of shady actors based in countries outside the US. It’s spam-adjacent stuff — if not outright spam.

This slop is sloppy enough that it can sometimes be very funny — and there’s even an X account, Insane Facebook AI Slop, dedicated to cataloging it.

It’s embarrassing to Facebook that people on Twitter/X — X!!! a place full of bad things!!! — are laughing at Meta’s poor content moderation.

This has been going on for quite a few months now. 404 Media, which has been diligently covering this, described how these images are part of a larger AI-fuelled “zombie internet” full of bots talking to bots. Not too long ago, I speculated about why Meta hasn’t cracked down, but considering nothing has seemed to change, I find myself wondering yet again …

Why doesn’t Facebook just ban this slop?

Here are a few of my theories:

• These images don’t technically violate any rules — no nudity or other forbidden imagery.
• Facebook did make a warning label to note an image might be made with AI when its systems detect it.
• Users actually enjoy this content — you see plenty of people in the comments who are probably duped and think it’s a real image, or maybe just an interesting artistic rendition. Why mess with content people like?
• Facebook does ban this stuff, but only at an account level when the account posting it engages in obvious spam tactics. But not all of the accounts are spamming.
• Slop is often scam-adjacent, but the accounts posting it may not be directly always the ones doing the scams, so they don’t get banned.
• Facebook worries that banning this kind of AI content will set a bad precedent because it wants to encourage people to make stuff with its own Llama 3 tools.
• Facebook doesn’t like the AI slop but has bigger things to worry about.
• Facebook has stopped trying. This is what Jason Koebler from 404 Media speculates. Meta’s previous commitments to working with researchers and academics have fizzled away, and the scrutiny over content moderation (helped by Elon Musk lowering the bar to the basement) has shifted. Meta is penny-pinching and focusing on the newest shiny thing (AI tools, Instagram, the Metaverse) and cares less and less about Facebook, this theory goes. Basically, they don’t give a shit anymore.

I imagine that some combination of all of these might be true to some degree. A representative for Meta didn’t reply to questions about why they don’t simply ban AI slop. (I had recently also asked a VP of product at Meta about why they don’t just ban this slop and he sort of changed the subject; it seemed like maybe he wasn’t sure what I was talking about.)

Facebook has gone through so many different flavors of spammy/scammy engagementbait, from Minions memes to videos from magicians, to actually dangerous fake news and extremist groups. If the AI slop isn’t suggesting you get together in real life and overthrow the government, hey, Facebook has seen way worse.