Let’s talk about the transaction she couldn’t complete.
She’s standing at her own bank, on her own phone, trying to move her own money. It’s not a huge amount. It’s not illegal. And it simply won’t go through. The app says a review is needed. The person on the phone says the system flagged it. When she asks flagged for what, the answer is a version of “we can’t share that.” She owns the money. She is not the one deciding what happens to it.
That story ends fine, and I’m not going to pretend banks haven’t stopped a lot of fraud with exactly this machinery. But notice what just happened in that exchange. A system made a judgement about her, applied it, and then stood in front of the reason. She wasn’t told what to do differently. She wasn’t offered the choice and asked to carry the risk. She was overruled.
That’s the thing I want to talk about. Not the fraud. The overruling.
Because the machine is getting extremely good at protecting you from yourself.
Look at where this is heading. Your car nudges you back into your lane and then starts taking the wheel. Your phone flags the link before you can click it. Your bank, your health app, your insurer, your AI agent — each one now holds a version of you and a list of things it has decided you shouldn’t be allowed to do. A lot of that is genuinely good. I’ll take the lane correction. I’ll take the scam warning.
But there’s a difference between a tool that hands you the odds and a system that cancels the bet.
And here’s what people are missing. The machine doesn’t protect you because it loves you. It protects you because it’s cheaper. When a system lets you do something stupid and you get hurt, somebody gets blamed, and it is almost never you. It’s the bank, the platform, the app. So the incentive is obvious and it only points one way: don’t let the person take the risk. Don’t build the button that says I understand the risk, let me proceed. Build the one that says this action isn’t available.
Nobody in that building thinks they’re being paternalistic, and they’re probably right about themselves. Protecting the person next to you has always been part of the price of living among people. The trouble is what happens when you scale that instinct and never have to defend it out loud.
Now slow down here, because this is the part that matters.
Judgement is not information. You can’t read your way into it. You can’t download it and you can’t inherit it from your parents. Judgement is what’s left over after you’ve made a call, watched it go badly, and had to live inside the result. The bad decision isn’t a failure of the process. It is the process.
Which means every time the machine steps in and lifts a bad decision off your hands, it doesn’t only save you a bruise. It removes one of the only things that was ever going to make you good at deciding.
We’re already running this experiment on grown adults. There are people in their twenties now who have never bought something a system didn’t pre-approve, never taken a route the app didn’t rank, never sent money to a person the fraud model hadn’t cleared, never once bet on themselves against a dashboard. Their lives are smoother than any generation before them. They’re also, in a specific and quiet way, less capable. Not because they’re weak. Because they’ve never been allowed to practise.
So I want to make one distinction, because without it this sounds like an argument for chaos, and I’m not making that argument.
There are two kinds of harm and we’ve quietly decided to treat them as one.
The first is harm you do to someone else. That has always belonged to other people. A community, a company, a court has every right to stop you from hurting the person beside you. That isn’t control. It’s the whole reason we agreed to live close together.
The second is harm you do to yourself. And that one has always been yours. It’s the zone where a person finds out what they’re made of, and where you learn the thing you can only learn one way, which is that you can survive a mistake and still be standing.
Safety, done badly, erases the second kind and calls it progress. It protects everyone from everything and quietly deletes the last room where a person was allowed to grow.
There’s a newer cost too. Your bad decision used to be small and local and forgettable. You did the dumb thing, you learned, and, mercifully, almost nobody ever heard about it. That’s mostly gone. Now the mistake is logged, scored, and priced into your future. The system learns what you do when you’re reckless and files it under who you are. So the penalty for the lesson keeps getting longer while the space to take it keeps getting smaller.
What do you do with this?
If you build things, stop designing tools that decide. Design ones that inform. Say the risk out loud, say it plainly, and then let the person carry it. That’s the harder product, and it’s the honest one. If your product can’t survive a user choosing badly with their eyes open, you didn’t build a tool. You built custody and you called it care.
And if you’re the one being protected, fight for one specific thing: the option that costs you something. The button that says you understand. The transaction that goes through. The route the app didn’t like. You won’t always be right. That’s the entire point. Being wrong on purpose and living through it is how a person stops being managed and starts being free.
Because the principle that has to survive here isn’t “safety is bad.” It’s narrower than that, and it’s older than any of this.
The right to make a bad decision. With your eyes open. In your own life. To be told the truth about what might happen, and then to be the one who decides anyway.
A machine can calculate the odds of your life going wrong. It can’t be the one who has to live with the outcome. And a person who never gets to live with an outcome never really becomes a person at all.
They just get managed into one.