Back to articles

Signal Limits Screenshots to Keep OS-Level AI Away From Chats

3 min read

Signal is restricting screenshots in its mobile apps in an effort to reduce the chance that operating-system AI agents can read private conversations from a device screen. The change responds to AI features that analyze on-screen information to build a better picture of a user. Signal argues that when system-level assistants can inspect app interfaces, an app’s own privacy protections may no longer be enough.

Key points

  • Signal says its apps on iOS 27 and Android will restrict screenshots to prevent assistants such as Siri and Gemini from reading visible content.
  • Users can restore screenshot functionality in Signal’s settings. The company warns that doing so may expose their own messages, as well as their contacts’ content, to AI agents.
  • Signal calls the measure incomplete: it cannot rule out other ways of accessing app data, including through a device’s file system.
  • The company says it cannot currently offer equivalent protections on macOS because developers lack adequate information and controls.

Signal president Meredith Whittaker frames the issue as a problem with the boundary between operating systems and apps. Developers, she says, do not have simple, robust tools for deciding what data AI agents can access. Signal is urging Apple, Google, and Microsoft to provide practical technical controls and accurate documentation. It also raises concerns about third-party agents using deep system permissions, such as accessibility features, to reach app information.

This is not Signal’s first attempt to block system-level data capture. The messaging service previously took steps to prevent Microsoft’s Recall feature from capturing protected chats. But AI agents may interact with device information through more than one route, and app developers may not be able to see or manage every access path. Signal presents screenshot blocking as a partial measure, not a guarantee that AI systems cannot obtain app data.

Why it matters

The move highlights a growing tension between encrypted messaging and AI features built into devices. End-to-end encryption protects messages in transit, but it does not by itself stop an assistant from interpreting content after it has appeared on a screen. Disabling screenshots adds another barrier, but it also removes a familiar app feature. Turning the restriction off restores convenience while potentially increasing exposure.

The underlying issue extends beyond Signal. As AI assistants become part of operating systems, app developers need clear ways to control screen-reading, accessibility access, and other routes to user data. Without predictable and verifiable controls, an app’s privacy promises can be affected by capabilities that its developers do not govern. Signal’s workaround is therefore also a call for platform makers to make AI access visible and manageable for both developers and users.

The Verge AI

Comments

Checking sign-in status...

Loading comments...

Related articles