Choose Your Character
From iOS to IDEs, the model-choice toggle is becoming a platform feature.
Apple and JetBrains, operating in entirely different markets, have both committed in recent weeks to letting users choose which AI model powers their platform’s core features. Google I/O begins tomorrow and Apple’s WWDC follows three weeks later, making the week ahead a useful moment to track a shift that is playing out across the industry.
Pick a provider
Bloomberg reported on May 5 that Apple plans to let users select a third-party AI model to power features across iOS 27, iPadOS 27, and macOS 27 through a system-level framework called Extensions. Siri, Writing Tools, and Image Playground will all route requests to whichever provider the user selects in Settings. Apple has been testing integrations with Google and Anthropic internally, alongside the existing ChatGPT option.
Until now, only ChatGPT had access to Apple Intelligence’s system-level integration, a position it has held since December 2024. Apple also signed a separate deal in January 2026 to use a custom Gemini model for native Siri features, which means that two distinct layers of third-party AI will operate simultaneously once Extensions launches this fall.
The mechanics mirror the existing default-browser setting. AI providers add Extensions support to their App Store apps. A user installs the app, selects it as the default in Settings, and the chosen model handles requests across all three surfaces. Siri will use distinct voices to signal which provider is responding, and Apple has posted a disclaimer stating that it bears no responsibility for outputs generated by external AI systems.
IDE plug-and-play
JetBrains published a strategic blog post in April 2026 declaring that its IDEs will not depend on a single AI vendor’s roadmap. The platform supports bring-your-own-key API access, OAuth sign-in to provider accounts, an agent connection protocol for external coding tools, and local model hosting through Ollama and LM Studio.
The stated rationale is blunt: the best model changes month to month, so the development environment must remain agnostic. Cursor already runs as an agent inside JetBrains IDEs through the agent connection protocol, demonstrating that competing AI tools can coexist in a single workspace.
Antitrust urgency
xAI filed an antitrust lawsuit against Apple and OpenAI in August 2025, alleging that ChatGPT’s exclusive Siri integration locked competitors like Grok out of the iPhone’s most valuable distribution channel. Extensions answers that claim directly by making the integration non-exclusive and user-selected.
The EU’s Digital Markets Act already requires Apple to offer browser choice screens and default app selection across the 27 member countries. The European Commission is now assessing whether AI assistants qualify as core platform services under the same law, which would subject them to identical contestability requirements. Apple’s Extensions framework could pre-empt that designation by demonstrating voluntary openness, much as Apple designed the browser choice screen to satisfy DMA scrutiny.
Samsung, meanwhile, is deepening Gemini integration across the Galaxy lineup with each generation, and Google’s model serves as the default intelligence layer for the platform. Apple’s and JetBrains’s model-choice layers double as competitive positioning against Google, which wants to own the default at every level of the stack.
Keynotes to watch
Google I/O begins tomorrow, with Gemini Intelligence, Googlebooks, and agentic features expected to dominate the keynote. Google has shown no indication of offering model choice on Android, and the keynote will clarify whether that position holds as competitors open their platforms.