For three years, before Brussels drafted a single clause, I sat in rooms where the entire agenda was one question: how do we make the off switch harder to find than the on switch. Nobody phrased it that bluntly. It came out as "reducing friction on re-engagement" or "optimizing the notification funnel." Same instruction, dressed for a board meeting. The default was never neutral. Someone chose it, tested it, shipped it, because the version where the user could leave in one tap converted worse than the version where they couldn't.
Nobody printed that part in the case study. The industry didn't stumble into attention capture. It selected for it, iteration after iteration, because the metric that funded the salaries was retention, and retention has never once been distinguishable, on a dashboard, from harassment.
Non-engagement was never a bug to fix
So when the European Parliament, inside the Digital Services Act framework and the Digital Fairness Act expected in the fourth quarter of 2026, starts using language like "the right to be left alone" as a digital right, that isn't a new idea arriving. That's an old debt being called.
Here's the distinction almost nobody draws correctly. For two decades, non-engagement got treated as a defect in the funnel. A user who didn't open the app, didn't respond to the push, didn't complete the flow, that user was a conversion failure to be corrected. A broken step in an otherwise working machine.
Not a person exercising a preference. A bug.
As I read it, this is the first time a legislative body of real weight treats non-attention and non-consumption as positive rights: not defects to patch, but states to protect. That is the reversal of that premise. The absence of a click stops being an error state. It becomes the outcome the design was supposed to permit all along.
Best practice and harassment share one file
I built systems on the other premise for years. I know exactly what a "frictionless unsubscribe" costs a P&L, because I ran the model. I know what happens to a 90-day retention curve when you make the off switch as visible as the on switch: it drops. Everyone in the room calls that a failure, because they're reading a chart that was never built to measure whether the user was better off. They built the dashboard to measure whether the user stayed. Those are not the same question. The industry spent a decade behaving as though they were.
Take the version of this that isn't a notification. A subscription box, say, a real one, not hypothetical. Cancellation buried four screens deep behind a chat widget that opens with "we'd hate to see you go" and a discount offer before it lets you finish the sentence you came to say. The company will tell you, and believe it when they say it, that this is retention best practice.
It is. That's exactly the problem. Best practice and harassment were never mutually exclusive. They were the same behavior, filed under different departments depending on who was asking.
Now put a second case beside it, because the pattern doesn't live in one industry. A news app reactivates a lapsed reader with a push notification worded as a question, "Still thinking about that story?", timed for 8:47pm because the data showed evening cortisol dips correlate with impulsive reopens. Nobody in that room called it manipulation. They called it a re-engagement campaign with a strong open rate. The vocabulary did the work of making the mechanism invisible. That's what vocabulary is for, in rooms like that.
The default itself becomes auditable
The European framework is attempting, and it is an attempt, not yet a settled fact, the Fairness Act is a proposal, not statute, to make the default itself the site of accountability. Not the message. Not the copy. The architecture underneath the message: is the exit engineered with the same care as the entrance. That's a question with an answer you can audit, which is exactly why it unsettles people who built their careers on the answer never being asked.
I've read the academic name for what sits underneath all this. The autonomy bias: the well-documented tendency of choice architecture to treat any non-choice as a problem requiring a nudge rather than a signal requiring respect. The regulation didn't invent the finding. It's catching up to something researchers have described for years while the industry kept building against it, because building against a known bias, at scale, paid better than building with it.
That's the difference between a norm and a preference.
A directive doesn't change what a designer wants, in the room, at 11pm, staring at a churn number that's red. It changes what they're allowed to ship without a paper trail.
That's not nothing. That's the difference between a norm and a preference. One gets enforced when nobody's watching. One doesn't.
Silence was the transaction all along
Here's the deeper claim I find buried in the legislative language, the one I think is worth sitting with longer than the news cycle allows: silence has a value that doesn't require justification. Not "the user was busy." Not "the user forgot." The user chose nothing, and nothing was the complete transaction. For a species of product design that has spent twenty years treating every unopened notification as an unsolved problem, that's not a policy update.
That's an admission.
Nobody builds an attention economy by accident and calls it a mistake once the invoice for the harm comes due. They call it product-market fit, until someone with the authority to write it into law calls it what it always was.
— Lucien Sael · Builder, Thinker, Operator
