Just when you’ve become grudgingly accustomed to being awash in AI slop every time you go online, along comes AI plop.
AI slop, to remind you, is the term given to that mass-produced, low-quality junk made using generative AI by idiots without an authentic bone or original thought in their bodies. Nobody wants it, but you’re getting it anyway, all over your social media. My current hatred is those short stories people are posting on Facebook, which purport to tell some sort of morality tale based on a celebrity’s philanthropic nature.
You’ll recognise the formula. Here’s an example: “[Celebrity Name] quietly built a $3m church in [location] for a struggling community. When asked why they didn’t publicise it, they said: ‘Real charity doesn’t need cameras.’ The church now serves hundreds of families every week.”
The posts all use the same story structure and just swap out the celebrity names to cater for the audience they’re trying to con. This can be accompanied by AI-generated images (presumably because charity does need AI) of the celebrity hanging out at the church. The stories often end with a heavy-handed, platitudinous moral. The most distressing thing about this slop is how many people share it, displaying the analytical skills and literacy of a small gerbil.
AI plop is… well, I’ll let Reddit user Ill_Car_7351 (not his real name) put it in his own words.
“I hoarded a large database of something valuable, just not what’s [sic] you expect … 150k stools images.” This Reddit post, which was brought to my attention by 404 Media, advertises exactly what it says — a database of poop images collected from PoopCheck, an AI poop-analysing app.
Around 25,000 people have taken pictures of their poop and uploaded them to the app. Ill_Car_7351 has been collecting, analysing and annotating these images and now wants to sell access to them.
“I’ve got 150k+ labelled and classified images of 💩 from roughly 25K different people. Jokes aside, I know there’s a lot of value in it (hard to obtain, useful for ML [machine learning] training, cancer studies etc) but not sure on how to move about it. Feels like I’m sitting on a pile of shi..ny coins but can’t find who wants them.” 404 Media adds that “the poster added that ‘the images are extremely rare’ and that he was trying to figure out how much money he could sell them for”.
The app that harvested these images is called PoopCheck, which takes you through a three-part process. First, open PoopCheck and take a quick photo of your bowel movement. “Our camera is optimised for fast, private capture — photos never leave your device unencrypted.” Second, get AI to analyse your stool. “Our AI engine classifies your stool using the Bristol stool scale, analyses colour, consistency, and compares against 140,000+ analyses worldwide.” And the third part is tracking your poop score. The AI apparently gives you a “daily gut health score” and lets you “spot trends”. What trends, I shudder to think.
I’ve written about technology writer Cory Doctorow’s concept of the “enshittification” of the internet, which he describes as a three-stage process of intentional platform decay driven by profit extraction. He coined the term in November 2022 to capture how digital platforms systematically degrade their services once they’ve captured their markets. I always thought that was a metaphor, but no. Turns out we’ve got a pretty literal example of it here with AI plop.
As humorous as this scatological romp through the bowels of the internet is, there’s a serious point to be made here. When it comes to feeding the gaping maw that is AI training data, privacy can often be brushed aside.
For all the positive uses of AI touted by its boosters, the reality is that, when you cut to the chase, AI is at heart a military tool used for mass surveillance
On its website, PoopCheck asserts that it is “committed to protecting your privacy” and says it’ll use your information “to provide AI stool analysis and gut health tracking services, to calculate and display your gut health score, to generate personalised health insights and recommendations, [and] to improve our AI analysis engine accuracy”.
It doesn’t mention anything about selling it to anyone who needs data to train AI. Confusingly, there is a business page that claims, “PoopCheck powers AI stool analysis for consumer apps, telehealth platforms, clinical tools, and research teams. Integrate our API, or license the labelled dataset it was built on.”
For all the positive uses of AI touted by its boosters, the reality is that, when you cut to the chase, AI is at heart a military tool used for mass surveillance. The UK National Cyber Security Centre says half of the world’s governments have access to commercial spyware that can break into computers and phones to steal sensitive information. This increases to 100 the number of countries with access to these types of hacking tools, up from the 80 countries estimated in 2023.
As much as people are being seduced by AI’s promise — and I count myself among them — it’s a very, very dangerous weapon. So, the fact that your personal data can become — and probably already has become — mere fertiliser to feed the beast should cause you some alarm.
And it’s a free-for-all when it comes to data being stolen. Recent data thefts include January’s breach of one of the most popular AI chatbot apps on Google Play and Apple App stores, Chat & Ask AI, which claims to have more than 50-million users.
The breach exposed 300-million private messages from more than 25-million users, revealing sensitive conversations, including requests about suicide. We’ve had a spate of cyber incidents in South Africa recently, with the Information Regulator reporting a 60% increase in data breaches between April and September 2025. The regulator told ITWeb that from January 1 to March 31, it received 788 data breach notifications from South African organisations. According to The Citizen, the country has the highest cyberattack rate globally, with 36% of organisations experiencing attacks over the past year.
The thing that really worries me, though, is how much more powerful state cyber surveillance becomes when it’s driven by AI. This is because it turns scattered data into actionable intelligence at scale. It’s one thing to have a data breach involving your passwords, because you can change those passwords. According to Breachsense, nearly one trillion stolen identity records are now circulating on the dark web.
But there’s also the breach of data that you can’t change. For example, the 6.9-million profiles exposed in 2023’s 23andMe breach, which included genetic ancestry data, names, birth years, family tree information, ancestry percentages and ethnic background. Not data you can really change if you’re trying to avoid surveillance.
It’s not just states that we have to worry about. In a recent article, TechCrunch describes how powerful iPhone hacking toolkits called Coruna and DarkSword, originally used by governments and advanced threat actors, have spread to cybercriminals and, in DarkSword’s case, partially leaked on GitHub in a plug‑and‑play form, creating a serious risk of silent data theft from any iPhone or iPad running older or unpatched iOS software.
“At least some parts of the Coruna toolkit were originally developed by Trenchant, a hacking and spyware unit within US defence contractor L3Harris, which sells exploits to the US government and its top allies. Kaspersky has also linked two exploits in Coruna’s toolkit to Operation Triangulation, a complex and likely government-led cyberattack allegedly carried out against Russian iPhone users.
“After Trenchant developed Coruna — somehow, it’s not clear how — these exploits found their way into the hands of Russian spies and Chinese cybercriminals, perhaps through one or several intermediaries who sell exploits on the underground market.” Yep. They’re coming for us from all sides.
With all this theft of our personal data, it seems quaint to worry about the fact that people’s faeces are being used to train AI. I can’t help thinking though, perhaps fancifully, that somewhere, some state agency is working on an AI solution called “Stool Pigeon”, which allows states to identify terrorists by their nervous bowel movements. The AI-driven enshittification of the internet carries on apace.
