Apple is drafting a framework that would let developers ship apps with autonomous AI agents while preserving the company’s strict privacy and code‑execution rules. The move follows a March crackdown that blocked updates to several “vibe coding” tools for letting users generate mini‑apps via natural‑language prompts, a practice Apple’s current policies deem unsafe for iPhone and iPad users.

Apple’s internal team is reportedly designing a sandbox that isolates AI‑driven code yet still allows features like flight booking or calendar invites to be handled by Siri and the forthcoming Apple Intelligence layer. Developers have been warned they may face commission fees later, though Apple says early‑stage integrations will be fee‑free. Talks with Baidu, Alibaba and Tencent have so far stalled over that point.

At WWDC on June 8 Apple will unveil the revamped Siri in iOS 27, which will support multiple third‑party chatbots—including models from Anthropic and Google—rather than relying solely on OpenAI’s ChatGPT. By opening the App Store to agentic AI while tightening execution controls, Apple hopes to curb rogue behavior that has plagued recent AI‑coding apps and keep its ecosystem secure across iPhone, iPad and Mac devices.