4 min read
Here is a sentence that should make your blood run cold, and it is not science fiction, it is a normal Tuesday on Instagram. Meta’s automated moderation is deleting the accounts, and with them the livelihoods, of thousands of innocent people, branding some of them with the worst accusation the platform has, and then handing their appeal to the very same AI that condemned them. If you are not famous enough for a journalist to make a few calls on your behalf, tough sh*t. That is the system now.
What is actually happening
Reporters at the New York Times, TechCrunch and the Guardian have documented a wave of wrongful bans across the US, Australia and New Zealand: a makeup artist with 48,000 followers, a disability advocate, a Juneteenth nonprofit, and, appallingly, people wrongly accused of sharing child-exploitation material, the single most destructive label Meta can put on a human being. Take Camille Hanson, an English teacher who spent years with her husband building a business teaching English to non-native speakers, close to a million followers. In March she woke to find both her Facebook and Instagram accounts flagged for deletion, the charge “fraud and deception”. A week later her appeal came back denied, with two sentences that belong in a dystopian novel: “All your information will be permanently deleted” and “You cannot request another review of this decision”.
The appeal is the same machine that convicted you
This is the part that should be illegal and somehow is not. Meta laid off huge numbers of the human reviewers who used to catch these mistakes, handed the job to AI to save money, and then handed the appeals to AI as well. So the algorithm accuses you, the algorithm judges your appeal, and there is no human anywhere in the loop you can reach, reason with, or shame. It is Kafka rebuilt as a cost-saving exercise. Computer says no, except the computer also runs the courtroom, and the sentence is the erasure of your business, your audience and years of your life.
Justice, but only if you are famous
Here is the detail that tells you exactly who this system is built to serve. Meta did restore some of these accounts, including one with nearly a million followers. When? After journalists intervened. Not after the appeal. Not after the evidence. After a reporter with a masthead behind them picked up the phone. Due process on the largest communication platform in human history has become a perk reserved for people big enough to embarrass Meta in print. Everyone else gets the two-sentence death notice and a permanently deleted life.
Meta’s defence, and why it misses the entire point
Meta’s response is that its newer AI tools make 13% fewer mistakes than human reviewers and catch 10% more violations, and that the accounts journalists surfaced were flagged by older systems. Fine. Grant them the numbers in full. Thirteen percent fewer mistakes at Meta’s scale is still millions of wrongly branded people, and “our robot is marginally less wrong than the humans we fired” is not the flex they seem to think it is. The problem was never only the error rate. It is that they built a machine to ruin people at industrial scale and then removed every human door a wrongly-accused person could knock on. Efficiency is not justice. It is just cheaper, and the difference is being paid by people who cannot afford a lawyer or a byline.
The AI tool stack actually worth paying for
One email a week. The tools, models and moves that matter, minus the hype and the horseshit filter set to maximum. Free.
Did you know: Meta moderates billions of accounts, a scale no court system on Earth could staff. That is precisely why handing both the verdict and the appeal to the same automated system is so dangerous: when the machine is wrong, there is no higher authority, no jury, and no appeal that is not simply the machine grading its own homework. Scale was the excuse for automating judgment, and it is also the reason the mistakes are uncountable.