
NEW YORK — Encrypted messaging service Signal is restricting screenshot access on supported mobile devices in a bid to stop artificial-intelligence agents built into phone operating systems from reading private conversations. The change is designed to prevent assistants that can interpret on-screen content from capturing messages and transferring them into other systems. Signal says the protections will apply to iOS 27 and to Android devices where the app can detect an AI assistant attempting to access the screen. On supported devices, the safeguards are enabled by default, although users can turn them off in the app’s settings if they want to take screenshots. Signal president Meredith Whittaker said the growing integration of AI agents into operating systems creates a challenge for applications that promise to protect sensitive information. In its official privacy statement published on November 20, Signal said screen-capturing agents could move information beyond the encrypted environment of the messaging app and into local databases or remote services controlled by other providers. Reuters reported the new restrictions and Signal’s request for stronger safeguards from Apple, Google and Microsoft.

The company has stressed that screenshot blocking is not a complete solution. It cannot prevent every possible route through which an AI agent might obtain information already stored on a device, including access through files or other system permissions. Signal also says it cannot yet provide equivalent protection on macOS because the operating system does not give developers reliable tools to block the relevant forms of screen capture. The company is asking platform makers to provide clearer documentation and practical controls that allow developers to opt out of AI access to screens, notifications and files without disabling useful features for customers. The dispute highlights a growing tension between the convenience of assistants that can see and act across apps and the privacy boundaries that individual services try to maintain. It also raises questions about what users understand when they grant an AI tool broad device permissions. Signal’s safeguards are a defensive step, not a guarantee that every form of automated data access has been eliminated. HeadlinesXtra has previously examined digital security risks in AI Cyberattacks Put Nigeria’s Digital Payments Network Under Pressure and the cyber intrusion that disrupted Japan’s Fugaku supercomputer.