OpenAI takes a swing at Apple's App Store — and devs should watch closely

What happened

OpenAI published a post titled "Apple is getting this wrong" — a rare move for a company that usually keeps its platform grievances behind closed doors. The piece lays out specific complaints about how Apple's App Store rules affect AI apps: opaque review decisions, forced use of Apple's in-app purchase system for AI subscriptions, and what OpenAI characterizes as preferential treatment for Apple Intelligence and its bundled ChatGPT integration versus standalone third-party AI apps.

The backdrop is important. Apple integrated ChatGPT into iOS 18 last year as an optional fallback for Siri, but that deal came with strings attached — Apple controls the surface, the routing, and the branding. Meanwhile, the standalone ChatGPT app on iOS competes against Apple's own AI features while paying Apple's 30% (or 15% for smaller developers) commission on Plus and Pro subscriptions sold through the App Store. OpenAI is essentially arguing that Apple gets to be both landlord and competitor, and it's using the landlord powers to squeeze the competitor.

The HN thread hit 86 points fast, which for a corporate-blog-post-as-complaint is unusually high. Comments split predictably: some see this as OpenAI whining after happily taking Apple's distribution for a year, others see it as a legitimate call-out that mirrors what Epic, Spotify, and Basecamp have been saying for years. The EU's Digital Markets Act looms over the whole discussion — Apple has already been forced to allow alternative app stores and payment systems in Europe, and the US antitrust case against Apple is still working its way through the courts.

Why it matters

The surface-level story is a platform-fee fight. The deeper story is about who owns the AI layer on a billion iPhones. Apple's play with Apple Intelligence is to become the operating system for AI — the router that decides which model handles which query, the identity layer that holds the user context, and the billing rail that captures the revenue. Every third-party AI app that runs on iOS is, from Apple's perspective, a candidate for absorption into that stack. OpenAI's blog post is a shot across the bow: don't assume we'll play along.

The economics matter for anyone building AI features. Apple's 30% cut on a $20/month ChatGPT Plus subscription is $6 — larger than the marginal inference cost for most queries. For a smaller AI startup running on OpenAI or Anthropic APIs, that 30% often exceeds gross margin entirely. The workaround — sell subscriptions on the web and let users log in on iOS — works for ChatGPT because the brand is strong enough that users will hunt down the web signup. It does not work for a new entrant whose first touch is the App Store.

There's also a review-process angle that developers underestimate until it bites them. AI apps sit in a category where Apple's guidelines are deliberately vague: "objectionable content," "minimum functionality," and the catch-all "we know it when we see it." When the reviewer is also your competitor, ambiguity is not neutral — it's leverage. Multiple AI app developers have reported rejections for features that Apple Intelligence itself ships (image generation, writing assistance, summarization) with no clear explanation of what would make the third-party version acceptable.

Community reaction on HN split between two grounded positions worth naming. The first: Apple's rules apply to everyone, OpenAI knew them going in, and "we don't like the tax" is not a policy argument. The second: platform gatekeepers with 50%+ US smartphone share and vertical integration into the exact market they regulate are the textbook definition of the problem antitrust law was built for, and OpenAI's willingness to say so publicly matters because most developers can't afford to.

What this means for your stack

If you ship an iOS app with LLM features, three things get more concrete this week. First, assume your monetization strategy needs a non-App-Store path. Web-first signup with a lightweight iOS client is now the default pattern for serious AI subscription businesses — Anthropic, OpenAI, Perplexity, and Cursor all do this, and the reason is Apple's tax, not a UX preference. If your funnel assumes in-app purchase, model what happens when 30% of your ARPU disappears.

Second, watch the DMA and the US antitrust ruling closely. If Apple is forced to allow third-party AI models as system-level defaults — the same way alternative browsers are now allowed in the EU — the entire competitive landscape for on-device AI shifts overnight. A world where users can set Claude or Gemini as their default assistant on iOS is a very different market than one where Apple Intelligence has the OS-level hooks and everyone else is a sandboxed app.

Third, if you're building against Apple's Foundation Models framework or the Writing Tools API, understand that you are building on a platform whose owner is actively fighting the company whose models you probably use in production. Apple's on-device models are getting better but are not competitive with frontier cloud models for most tasks. The temptation to lean on Apple's APIs for the free inference and OS integration is real — just don't design yourself into a corner where a platform-policy change breaks your product.

Looking ahead

This blog post will not, by itself, change anything. But it marks the moment OpenAI decided its relationship with Apple is adversarial enough to say so in public, and that shift matters. Expect more AI companies to follow — Anthropic, Perplexity, and Mistral have all had quiet complaints about App Store reviews, and OpenAI going first lowers the cost of the second and third company doing the same. The regulators are already circling; the question now is whether developers will organize their own pressure, or wait to inherit whatever settlement the lawyers deliver in 2027.