Apple, the EU, and the Cost You Don't Normally Get to See

Sixteen years ago, in the early days of the App Store, we built an app called The Red Carpet. It let you follow the Oscar nominees, wrapped around an audacious hook: guess every winner correctly and win a million dollars. The contest was real. We’d secured a $1M policy from Lloyd’s of London, gotten bonded across multiple states, and done all the unglamorous legal work a real contest requires. Then we submitted it, well ahead of Oscar season, and waited.

Apple rejected it. The reason given was that they don’t allow contests. So we went back through the developer agreement and the guidelines looking for the rule we’d broken. There wasn’t one. There was just a policy, unwritten and apparently absolute, and a phone call telling us there was no appeal. We were now carrying an expensive insurance policy for an app nobody was allowed to see.

That is what foreclosure looks like from the inside. Not a “no” you can argue with, but a wall you only find after you’ve already spent the money. Apple, to their credit, eventually listened, rewrote the guideline to allow properly-backed contests, and the app went live. But for a good while, the thing we’d built simply didn’t exist as far as the world could tell.

The cost you can’t see

When a rule keeps something from being built, the thing that never gets built leaves nothing behind. There is nothing to point at: no shipped product, no missing feature, no user to ask what they lost. Our app was a strange exception only because it already existed when it got foreclosed. The far more common case is the app that never gets written at all, because amid the uncertainty, a company decides the risk isn’t worth it and quietly moves on. Multiply our one rejection by every developer in those early years who saw how the rules actually worked and chose not to gamble, and you get a substantial graveyard of software no one will ever see, because it never reached the point of existing.

This is what makes “just regulate it away” such an easy argument to win. The benefits of a restriction are visible and immediate. The costs are invisible and deferred. You can’t weigh a tradeoff you can’t see, so we mostly pretend the tradeoff isn’t there.

This time, we get to see it

Which is why Apple’s standoff with the European Union presents an interesting lens: here’s a case where a foreclosed feature is still fully developed and visible.

Apple’s new AI version of Siri is shipping this year in the United States and nearly everywhere else. It is not shipping in the European Union. The reason, stripped of the press releases on both sides, is the Digital Markets Act’s interoperability requirement: if Siri AI can read your messages, move through your apps, and act on your behalf, then under the DMA any competing assistant has to be granted that same deep access the moment Apple ships. Apple says they cannot do that safely, so they are choosing not to ship the feature in Europe at all.

John Gruber, and even the Washington Post editorial board, have already made the case that this is bad policy, and I agree with them. But the part I find most useful is the one they mostly pass over. This time, the foreclosed thing is visible. The feature exists. Americans will use it. Europeans, in all likelihood, will not. We are about to run a natural experiment in what a regulation actually costs: a working feature in one market, a blank space where it would be in the next. The road not taken is usually hidden. This time we can walk up and look at it.

The EU has a point

The Commission is technically correct when it says nothing in the DMA forbids Apple from launching Siri AI. What the law forbids is the version Apple built. The Commission’s goal is contestability: it does not want Apple using control of the operating system to lock every user into Apple’s own assistant by default. The same contestability desire has guided the decisions around browsers and payment mechanisms.

But the DMA looks at Apple the way fantasy sports looks at athletes: take the parts that perform, swap in whoever you like, sum the stats, and the result should be at least as good. As a Lakers fan, having watched the LeBron/Westbrook era, I can tell you how painfully wrong that is. There might be two MVPs, but there’s only one ball. A company, like a team, is more than the measurable stats in a spreadsheet. Especially for Apple, the integration is the product, and assuming that parts can be swapped in and out without changing anything ignores reality.

The security risk is where the two sides simply will not meet. To comply, Apple would have to hand any third party the same intimate, autonomous access to your device that they grant their own assistant. Security researchers have already shown how AI agents with that reach can be hijacked to exfiltrate data and quietly alter a user’s files and settings. It’s actually a pretty bold move for Apple to be shipping in the first place: one big story of how an innocent text message from a stranger allowed an attacker to steal someone’s bank information and suddenly Siri AI casts doubt on all of iOS. But Apple certainly isn’t going to just hand that power out because EU regulators want them to.

Apple had it coming

None of which lets Apple off the hook, because Apple spent years teaching regulators to distrust them by using self-serving arguments to protect their business. The clearest public example is steering. Apple wouldn’t let apps so much as link out to their own websites if the website contained any way to pay. You could download Netflix or Spotify and find no way to buy anything inside, because Apple wouldn’t allow a sentence pointing you elsewhere. That didn’t protect users, it confused them. App Store reviewers were told to hunt down every possible construed violation. Apps were rejected because you could click on their privacy policy, poke around the web view ever so craftily and somehow land at the membership page.

The tragedy of it all was that Apple’s payment system was actually good with a ton of built-in advantages. It’s safe, users trust it thoroughly, they already have their credit card on file, they can pay with their thumb or their face and they know they can cancel without a hassle. Apple could have easily allowed apps to link out, putting the user burden back on the app makers. I know plenty of people (myself included) who would have still used Apple’s payment methods. Instead, Apple pretended that secure payments on the web don’t exist and only the App Store can take a credit card.

So when the EU decided a company controlling access to half a billion EU users couldn’t be trusted to set the terms by itself, it wasn’t working from nothing. Apple built that impression one capricious and self-serving rejection at a time. The uncomfortable symmetry is that the DMA is now doing to Apple roughly what Apple did to developers: build the whole thing, then find out afterward whether you’re allowed to ship it.

What we finally get to see

So the EU and Apple end up both harming each other. But the ones who pay the real costs are users. European iPhone owners will go without a feature the rest of the world has. In the App Store years, it was users who downloaded apps they couldn’t find a way to pay for, and developers who ate the cost of building things that died on submission. The burden also falls unevenly: a company Apple’s size can absorb the cost of complying, or of walking away, in a way a small team never could. It’s a large part of why GDPR, for all its intentions, ended up entrenching the very largest players while raising the bar for everyone trying to compete with them. The unbuilt thing is rarely the giant’s loss. But everyone is poorer for it, because we all live in the world where the better thing never got made.

This is the part that should matter to anyone who builds software for a living, regulators included. There is a deep difference between a rule you can build against and a verdict you can only receive. A specification you can check tells you, before you spend the money, whether what you’re building is allowed. A judgment rendered after the fact tells you only that you guessed wrong. Apple ran their store on the second model for years; the EU is running the DMA on it now. The result is the same in both cases: people stop building the things they can’t be sure will survive contact with the gatekeeper.

We rarely get to see the cost of that. This time, sitting right there on the other side of a border, we do. If Siri AI turns out to be as good as Apple thinks it is, the tradeoff will be undeniably visible.