TL;DR
Apple will reportedly allow users to choose which third-party AI models power features in iOS 27, turning the operating system into a platform for competing AI assistants. This shift, expected to be announced at WWDC 2026 in June, marks a radical departure from Apple's historically closed ecosystem and could reshape the mobile AI market.
What Happened
Apple plans to make iOS 27 a "Choose Your Own Adventure" for AI models, according to a TechCrunch report published Tuesday, May 5, 2026. Users will reportedly be able to select from a menu of third-party large language models—including options from Google, Anthropic, Meta, and potentially OpenAI—to handle tasks like text generation, image creation, and voice assistance directly within the operating system.
Key Facts
- iOS 27 is expected to launch in September 2026, with a developer beta arriving at WWDC in June 2026.
- Users will select their preferred AI model via a system-level settings panel, similar to how iOS currently handles default browser or email apps.
- Apple has held talks with Google (Gemini), Anthropic (Claude), Meta (Llama), and OpenAI (GPT-5) about integration, according to sources familiar with the discussions.
- The move follows Apple Intelligence, Apple's own on-device AI system introduced in iOS 18 in 2024, which handled tasks like summarization and image editing without third-party models.
- Privacy remains a key concern: Apple is reportedly building a secure enclave for model inference that prevents third-party AI providers from accessing user data directly.
- The feature is expected to be opt-in, with Apple's own AI models remaining the default for privacy-sensitive tasks like health data processing.
- Analysts at Morgan Stanley estimate this could generate $5–8 billion annually in revenue-sharing fees from AI providers by 2028, based on user adoption rates of 30–40%.
Breaking It Down
Apple's decision to open iOS to third-party AI models represents a fundamental strategic pivot. For over a decade, the company has tightly controlled which services and apps can access core system functions—Siri, for instance, has only recently allowed limited third-party integrations. By turning iOS into a platform for competing AI models, Apple is effectively acknowledging that no single AI provider can satisfy all user needs, and that the real value lies not in the AI itself but in the distribution and trust Apple commands.
40% of iPhone users surveyed by Consumer Intelligence Research Partners in April 2026 said they would "likely or very likely" switch to a third-party AI model if given the choice, with speed and accuracy cited as primary motivators.
This statistic underscores the risk Apple is taking. If users overwhelmingly abandon Apple's own AI for third-party alternatives, the company risks losing the data flywheel that powers its on-device machine learning improvements. However, Apple appears to be betting that its privacy infrastructure—specifically the Secure Enclave for AI—will become the differentiator. Unlike Android, where Google's Gemini can access vast amounts of user data, Apple's system will reportedly sandbox third-party models so they only receive anonymized, task-specific inputs.
The revenue implications are significant but complex. Apple traditionally takes a 15–30% cut of in-app purchases and subscriptions through its App Store. For AI models, the model is still unclear: providers might pay Apple a per-user licensing fee, a share of subscription revenue, or a flat fee for placement in the settings menu. The Morgan Stanley estimate of $5–8 billion annually assumes a 20% revenue share on AI subscriptions, which currently range from $10–$30 per month for premium tiers from companies like OpenAI and Anthropic.
What Comes Next
- WWDC 2026 (June 8–12, 2026): Apple is expected to formally announce the iOS 27 AI model selection feature, along with the list of launch partners. Developers will gain access to the APIs needed to integrate their models.
- Developer beta (July 2026): The first beta of iOS 27 will let developers test the AI model switching mechanism. Expect early benchmarks comparing Apple's on-device model against third-party cloud-based models for speed and accuracy.
- Regulatory scrutiny (late 2026): The European Commission's Digital Markets Act already forced Apple to allow alternative app stores and browser engines. This AI model choice feature could face similar demands for equal treatment of smaller AI providers, potentially delaying or altering the rollout.
- Revenue-sharing negotiations (ongoing through 2027): Apple is reportedly seeking 20–25% revenue share from AI providers, but some companies—particularly OpenAI—are pushing for a flat annual fee instead. The outcome will set a precedent for the entire mobile AI economy.
The Bigger Picture
This story sits at the intersection of two major trends: Platformization of AI and Regulatory pressure on mobile ecosystems. The Platformization of AI trend sees major tech companies—Apple, Google, Microsoft, Meta—racing to embed generative AI into every layer of their operating systems. Apple's approach differs by making its platform a curated marketplace for AI models, rather than building a single, dominant model. This mirrors the App Store strategy that turned iOS into the most profitable mobile ecosystem.
The Regulatory pressure on mobile ecosystems trend, driven by the EU's Digital Markets Act and similar legislation in the UK, Japan, and India, has forced Apple to open its walled garden. The AI model choice feature can be seen as a preemptive move to avoid future mandates requiring Apple to allow any AI model onto its devices. By setting the terms now—privacy guarantees, revenue sharing, and user opt-in—Apple hopes to shape regulation rather than react to it.
Key Takeaways
- [Strategic Pivot]: Apple is abandoning its closed AI approach by letting users choose third-party models, acknowledging that no single AI provider can dominate all use cases.
- [Privacy as Moat]: Apple's Secure Enclave for AI inference is designed to prevent third-party models from accessing raw user data, positioning privacy as the key differentiator against Android.
- [Revenue Opportunity]: Analysts project $5–8 billion in annual revenue by 2028 from AI model revenue-sharing, though negotiations with providers are ongoing and contentious.
- [Regulatory Hedge]: The feature preempts potential DMA-style mandates by giving Apple control over which models are available and under what terms, rather than being forced to allow all models.
