Apple bans vibe‑coding apps, devs push back

iPhone apps are supposed to arrive as finished packages. That is the whole App Store deal — Apple reviews the software, approves it, and then users run what Apple checked. But AI “vibe-coding” tools break that model. They let people type a prompt, generate a new app or feature on the fly, and often preview it right inside the iPhone app. In March and April 2026, Apple started pushing back hard, blocking updates from apps like Replit and Vibecode and removing the app Anything from the App Store. ### What is a vibe-coding app? A vibe-coding app is basically an AI builder for non-programmers. You describe what you want in plain English, and the tool generates software — a game, a utility, a website, sometimes even a mobile app prototype. The important part is not just code generation. It is that the app often lets users run or preview that generated software immediately, inside the same container Apple approved. ### Why does Apple care so much? Apple’s objection is not really “AI made this.” The objection is “this app can become a different app after review.” Rule 2.5.2 says apps must be self-contained and may not “download, install, or execute code” that changes the app’s features or functionality. Apple also points to its developer license, which allows interpreted code only if it does not change the app’s primary purpose. That is the line Apple says some of these tools crossed. ### What actually happened? The first visible move was Apple blocking updates to Replit and Vibecode unless they changed how their apps worked. Then Apple escalated. Anything — a vibe-coding app built by Dhruv Amin — was removed from the App Store after Apple rejected an update that was meant to address the issue. Apple’s position, repeated to reporters, was that this was enforcement of longstanding rules, not a brand-new policy invented for AI tools. ### Why is the preview the flashpoint? Because the preview is where generated code stops being text and starts behaving like software. If an app creates a new tool and runs it inside an embedded web view or similar container, Apple can argue that the shipped app is now executing fresh, unreviewed functionality. From Apple’s perspective, that looks a lot like sideloading a mini-app inside an approved app. From the developer perspective, that is the whole point of the product. ### Didn’t Apple already allow code in some apps? Yes — but narrowly. Apple’s own guideline carves out educational apps that teach or let students test executable code, as long as the code is fully viewable and editable by the user and not used for other purposes. That exception fits coding tutors better than consumer AI app factories. So developers arguing “but coding apps exist already” are running into a rule that was written for learning environments, not one-click app generation. ### Why are developers so mad? Because this feels bigger than one guideline fight. Developers see Apple embracing AI-assisted coding in Xcode while limiting third-party tools that could let ordinary users build software without going through traditional app-development funnels. The complaint is not just inconsistency. It is that Apple’s review model may be fundamentally incompatible with software that is supposed to change in response to a prompt. ### So what happens now? The obvious workaround is to push the generated app preview into Safari or the open web instead of running it inside the iPhone app. That may satisfy Apple in some cases, but the catch is that it makes the experience clunkier and weakens the promise of “build an app on your phone, instantly.” Some developers may also look harder at alternative distribution in places where Apple now allows it, like the EU. ### Bottom line? This is really a fight over what an app is. Apple wants reviewed, bounded software. Vibe-coding tools want software that keeps rewriting itself. Both ideas can work — but on iOS, they are starting to look mutually exclusive.