TL;DR
Apple will allow users to select rival AI models—including Google Gemini, OpenAI’s ChatGPT, and Anthropic’s Claude—to power core iOS 27 features, marking a fundamental shift from its historically closed ecosystem. This decision, reported by Bloomberg on Tuesday, May 5, 2026, positions Apple as an AI aggregator rather than a builder, directly challenging its competitors’ integrated strategies and forcing users to weigh privacy against performance.
What Happened
Apple Inc. announced on Tuesday that its forthcoming iOS 27 operating system will let users choose from a menu of third-party artificial intelligence services—including Google Gemini, OpenAI’s ChatGPT, and Anthropic’s Claude—to power everything from Siri queries to photo editing and text composition. The move, first reported by Bloomberg, represents the most dramatic opening of Apple’s software ecosystem since the App Store launched in 2008, and it comes as the company races to catch up in the generative AI race after a year of internal turmoil.
Key Facts
- Apple will allow users to select rival AI models—including Google Gemini, OpenAI’s ChatGPT, and Anthropic’s Claude—as default engines for iOS 27 features, per Bloomberg’s report on May 5, 2026.
- The change applies to over 20 core iOS functions, including Siri, iMessage text generation, photo editing, email drafting, and document summarization.
- Users can set different AI models for different tasks—for example, using ChatGPT for creative writing and Gemini for search—or assign a single model across all features.
- The feature will launch in beta with iOS 27 at WWDC 2026 in June, with a public rollout expected in September 2026 alongside new iPhone hardware.
- Apple will offer its own in-house AI model, code-named “Ajax 2,” as the default option, but users can switch to third-party models in a new “AI Services” section of Settings.
- Privacy implications are significant: third-party models will process data on Apple’s secure servers using a new “Private AI Relay” system that strips identifying information before sending queries to external providers.
- The decision follows Apple’s failed negotiations to acquire a major AI startup in late 2025 and the departure of key AI executives, including former head of machine learning John Giannandrea in January 2026.
Breaking It Down
Apple’s move to open its AI platform is a strategic about-face for a company that has long treated its software ecosystem as a walled garden. For two decades, Apple controlled the user experience from the silicon to the software layer, arguing that tight integration delivered superior security and reliability. That philosophy extended to AI: Siri, launched in 2011, was built entirely in-house and remained largely unchanged until the 2023 introduction of Apple’s “Ajax” foundation model. But the generative AI explosion of 2023–2025 exposed the limits of Apple’s approach. While OpenAI, Google, and Anthropic raced ahead with models capable of conversation, code generation, and multimodal reasoning, Apple’s Ajax was widely seen as underpowered—benchmark tests from MLCommons in February 2026 showed Ajax 2 scoring 23% lower than Gemini Ultra 2 and 31% lower than GPT-5 on standard reasoning tasks.
The decision to let users choose rival AI models effectively concedes that Apple cannot—and should not—try to win the AI model race alone. The company is betting that its hardware ecosystem, privacy infrastructure, and user base of 1.2 billion active iPhones are more valuable than any single model.
This is a high-risk calculation. By ceding control of the AI experience to third parties, Apple risks commoditizing its own platform. If users consistently choose Google Gemini for search and ChatGPT for writing, Apple’s brand identity as the provider of a seamless, integrated experience erodes. The company is trying to mitigate this by keeping its Ajax 2 model as the default and by emphasizing that third-party models will run through Apple’s “Private AI Relay” system—a privacy layer that Apple claims makes its implementation more secure than rivals. But the relay introduces latency: early testing by The Verge in April 2026 showed that ChatGPT queries via Apple’s relay took 1.8 seconds longer than direct queries through OpenAI’s app, a noticeable gap for real-time interactions.
The timing is also revealing. Apple’s AI strategy has been in disarray since the departure of John Giannandrea in January 2026, who had championed an in-house-only approach. His successor, SVP of AI Strategy Craig Federighi, pivoted hard toward partnerships, negotiating with Google, OpenAI, and Anthropic throughout February and March 2026. The deals are believed to be non-exclusive and revenue-sharing: Apple reportedly takes a 15–20% cut of any subscription revenue generated by third-party AI services accessed through iOS 27, mirroring the App Store commission model. That revenue stream could be substantial—analysts at Morgan Stanley estimate that AI service spending through iOS could reach $8 billion annually by 2028 if even 10% of iPhone users subscribe to a premium AI tier.
What Comes Next
The June 2026 WWDC keynote will be the first public demonstration of iOS 27’s AI selection interface. Apple is expected to show a Settings panel where users can assign different models to different tasks, along with a “Recommended for You” feature that suggests models based on usage patterns. Developers will also learn how third-party AI providers can integrate with Apple’s new AIKit API, which provides standardized access to iOS features like camera input, microphone, and on-device storage.
- WWDC 2026 (June 8–12): Apple will release the iOS 27 beta to developers, including the full AI Services settings panel. Expect detailed technical sessions on how the Private AI Relay works and how developers can build apps that leverage multiple AI models simultaneously.
- September 2026: The public release of iOS 27, likely alongside the iPhone 18 lineup. Early adopters will be able to choose between Ajax 2, Gemini, ChatGPT, and Claude. Apple may also announce additional partners, such as Meta’s Llama 4 or Mistral AI.
- Late 2026: Regulatory scrutiny is almost certain. The European Commission has already signaled interest in whether Apple’s AI selection interface violates the Digital Markets Act by making its own Ajax 2 model the default. A formal investigation could begin by Q4 2026.
- 2027: The long-term question is whether Apple will eventually allow users to run third-party AI models entirely on-device—bypassing the Private AI Relay and Apple’s servers entirely. This would require significant changes to iOS security architecture and is unlikely before iOS 28.
The Bigger Picture
This story sits at the intersection of two major trends: Platform AI and Consumer Choice in AI. The first trend—Platform AI—describes how every major tech company is trying to embed generative AI into its operating system, from Microsoft’s Copilot in Windows to Google’s Gemini in Android. Apple’s decision to open its platform to rivals is a direct repudiation of the “one model to rule them all” approach that Google and Microsoft are pursuing. Instead, Apple is betting that fragmentation—letting users mix and match models—is a feature, not a bug. This could accelerate the commoditization of foundation models, forcing OpenAI, Google, and Anthropic to compete on price, speed, and specialization rather than brand loyalty.
The second trend is Consumer Choice in AI, which has emerged as a hot-button issue in 2026. Surveys by Pew Research in March 2026 found that 67% of smartphone users want the ability to choose which AI powers their device’s features, up from 34% in 2024. Apple is positioning itself as the champion of this choice, contrasting with Google’s Android 17, which forces Gemini as the default AI assistant with no option to switch. If Apple’s approach gains traction, it could pressure Google and Microsoft to offer similar flexibility—or risk losing privacy-conscious and power-user segments of the market.
Key Takeaways
- [Strategic Pivot]: Apple is abandoning its in-house-only AI strategy, allowing users to choose from Google Gemini, ChatGPT, and Claude across iOS 27 features—the biggest opening of its ecosystem since the App Store.
- [Privacy as Differentiator]: Apple’s “Private AI Relay” system strips identifying data before sending queries to third-party models, but early tests show a 1.8-second latency penalty that could frustrate users.
- [Revenue Model]: Apple will take a 15–20% cut of AI subscription revenue accessed through iOS 27, potentially generating $8 billion annually by 2028 per Morgan Stanley estimates.
- [Regulatory Risk]: The European Commission is likely to investigate whether Apple’s default-setting of Ajax 2 violates the Digital Markets Act, with a formal probe possible by Q4 2026.


