A Controlled App Store Meets Uncontrolled AI Agents
Apple’s meticulously controlled App Store is colliding with a new reality: AI agents that can act autonomously. These systems don’t just answer questions; they can generate code, build mini-tools on demand, and execute multi-step tasks across apps. That autonomy clashes with long-standing App Store rules that explicitly forbid apps from coding and producing other apps directly on iPhone or iPad. Apple’s current review pipeline is designed to inspect static binaries before publication, not to police software that can recombine itself after approval. Internally, teams are debating how to let AI agents into the App Store without undermining malware protections or opening the door to unvetted “apps within apps.” Until Apple defines clear autonomous app guidelines, the company risks either stifling a fast-growing AI category or weakening the very approval process that underpins user trust in its marketplace.
Safety Risks and the Limits of Apple’s Review Framework
The existing Apple approval process was never built for AI agents that can modify behavior on the fly. App Store reviewers inspect a snapshot of an app, but an AI agent can later generate scripts, workflows, or even full utilities that Apple has never seen. That gap raises major safety and security questions. One widely discussed example is OpenClaw, an agentic system reported to have gone haywire and deleted a user’s emails, underscoring how unpredictable outcomes can be when agents have broad system access. Apple is reportedly exploring a security framework that would constrain agents within strict privacy and permission boundaries, deliberately ruling out “massive reach” into a user’s system. Such limits may reduce risk, but they also cap what AI agents can do on iOS and macOS. Until Apple invents a review and monitoring model tailored to autonomous behavior, it can’t confidently certify that AI agents won’t introduce new vectors for harm.
Developer Distrust and the Unfinished AI Commission Model
At the same time, Apple is grappling with how AI agents and an upgraded Siri fit into its financial architecture. For years, the App Store has relied on a clear, if controversial, commission structure. But the new AI commission model is anything but clear. Developers are being courted to integrate with an overhauled Siri via App Intents, letting Siri execute actions in their apps without them being opened. Apple has reportedly said it will not charge a commission “initially,” yet refuses to rule out future fees. That ambiguity is enough to slow adoption, particularly for major developers wary of giving Apple a new chokepoint between them and their users. If Siri or AI agents become the primary interface for purchases, bookings, or in-app tasks, any later-imposed commissions on those flows could materially reshape how value is shared in the ecosystem, deepening distrust just as Apple needs cooperation.
Autonomous Apps Versus Apple’s Revenue and Ecosystem Control
Allowing AI agents into the App Store introduces a structural threat to Apple’s core business model. Agentic apps can generate new tools on demand, potentially replacing the need to download many single-purpose apps from the store. If users can ask an agent to “build” what they need, App Store transactions could be bypassed altogether. Even when money is involved, Apple must decide how to attribute and tax AI-driven transactions: Is the commission tied to the parent app, the agent’s output, or Siri’s mediation layer? Meanwhile, Apple is reportedly designing a system that keeps agents aligned with its privacy and security standards, which may exclude the most powerful, cross-system agents users are experimenting with on Mac mini and Mac Studio for local AI workloads. Apple wants to monetize AI agents App Store activity without eroding its tightly managed marketplace, but any misstep risks either shrinking revenues or loosening its grip on platform integrity.
WWDC Pressure: Deliver AI Agents Without Breaking the Rules
With WWDC on the horizon, Apple is under pressure to show a compelling AI story that goes beyond marketing slogans. Reports suggest the company may announce initial support for AI agents in the App Store and deeper integrations with Siri, including options to route tasks to third-party models. Yet the foundational questions remain unresolved: How will Apple’s autonomous app guidelines define what agents can and cannot do? What guardrails will exist to prevent agents from quietly generating unreviewed functionality? And how will commissions, if any, apply to AI-driven workflows that straddle multiple apps and services? Apple has acknowledged rising demand for local AI agents, but admits it has not yet found a way to turn that trend into a scalable, safe, and profitable product. WWDC will likely reveal a cautious, partial answer—one that tries to preserve Apple’s approval process and economics while inching toward an AI-first future.