Android 16 QPR3 Brings ‘Screen Automation’ to Pixel 10
▼ Summary
– Gemini’s “Computer Use” feature is currently available for desktop web users subscribed to AI Ultra, with Android availability being prepared.
– Code analysis of an Android beta version reveals a new “Screen automation” system permission that would allow apps to interact with other apps’ screen content.
– The Google app, which powers Gemini, is currently the only app listed to use this permission, with options to always allow, ask, or deny access.
– This feature, internally called “computer_control,” is designed to let an AI agent help complete tasks by clicking, typing, and scrolling like a human would.
– Google demonstrated similar on-device AI capabilities in May, but the public release date for Gemini’s Android computer control remains uncertain.
The latest Android 16 QPR3 Beta 2 introduces a significant new permission called “Screen automation,” hinting at a major expansion of Google’s Gemini AI capabilities directly onto Pixel smartphones. Currently, the “Computer Use” feature for Gemini is limited to desktop web browsing for AI Ultra subscribers, but this discovery within the system settings strongly suggests a mobile version is in active development. This permission, which appears exclusively on Pixel 10 devices in the beta, would grant approved applications the ability to view and interact with the content on other apps’ screens to assist users with automated tasks.
Found under Settings > Apps > Special app access, the Screen automation entry includes the description: “Allow apps to help you complete tasks by interacting with other apps’ screen content.” At present, the Google app, which powers Gemini, is listed as the sole application able to request this powerful access. Users are presented with three control options: Always allow, Ask every time (which is the default setting), and Don’t allow.
The underlying code provides more context about how this feature would operate. One string explicitly states that an app with this permission “will be able to see and interact with other apps’ screen content to help you complete tasks, even when the apps are in the background.” This points toward a system where an AI agent could perform multi-step processes across different applications without constant user supervision. Additional code references, labeled “computer_control,” describe functionality for launching apps automatically and warning users about stopping tasks mid-progress to avoid losing their place.
This development aligns with Google’s demonstrated vision for AI agents that navigate digital interfaces “just as humans do: by clicking, typing and scrolling.” Earlier demonstrations, like those from Project Astra, showed prototype AI seamlessly scrolling through Chrome on Android and interacting with the YouTube app. The Screen automation permission in Android 16 QPR3 is clearly the foundational system-level groundwork required to bring those ambitious “computer use” demonstrations to a live consumer product.
While the infrastructure is being put in place, it is not yet active for end-users. The timing for when Gemini will actually utilize this Screen automation capability on Pixel phones remains uncertain. Furthermore, its current limitation to the Pixel 10 series in the beta raises questions about whether older devices, like the Pixel 9, will eventually gain support or if this represents a hardware-dependent feature reserved for newer models. This move signifies Google’s intent to deeply integrate advanced, agent-like AI directly into the Android operating system, potentially transforming how users accomplish complex tasks on their devices.
(Source: 9to5 Google)
