Signal is extending screenshot restrictions to its Android and iOS apps, as the encrypted messaging service warns that operating-system-level artificial intelligence could expose information displayed in private conversations. Signal President Meredith Whittaker called the growth of agentic AI an “existential threat,” arguing that tools able to scan what is on a user’s screen can put sensitive material beyond the app’s control.

The change follows Signal’s existing approach on Windows, where screenshots are blocked by default. The new mobile restrictions are optional: users can turn the setting off, but Whittaker said doing so would “break our privacy guarantees.” The announcement reflects a widening concern for privacy-focused software makers: protecting data inside an app may not be enough if another layer of the device can inspect what the app displays.

What Signal is changing

Signal said Friday that it is adding screenshot limits to its Android and iOS apps, bringing those mobile platforms closer to the protection already applied by default to its Windows app. The measure is intended to make it harder to capture conversation content through ordinary screenshots. Signal has not presented the change as a complete defence against every way information could be accessed on a device.

Whittaker’s concern is that AI agents built into operating systems may be granted the ability to read or scan screen contents to carry out tasks for users. She pointed to Windows, Android and Apple’s macOS and iOS as platforms where operating-system-level AI could, in her view, undermine Signal’s ability to independently guarantee privacy. She said captured information could be stored or made available to a local database or remote server used by an AI agent.

Those statements describe the risk Signal sees, rather than a report that a particular agent has accessed Signal conversations. The distinction matters: the announcement is a precautionary response to the way operating-system features may interact with private app data. Signal says it cannot control what happens after information displayed on screen has been captured outside the app.

A safeguard with limits

Blocking screenshots addresses one visible route for copying information, but it cannot by itself prevent every form of screen access. Whittaker acknowledged that the measure is “incomplete” and does not address other, less obvious ways an AI agent might potentially obtain application data. The company’s own description therefore frames the setting as a limited safeguard, not a guarantee that all information on a device remains inaccessible to other software.

The optional nature of the setting also leaves users with a practical choice. Screenshots can be useful for saving information or sharing part of a conversation, but Signal’s warning makes clear that disabling the restriction weakens the privacy protections the app aims to provide. Users will need to weigh that convenience against the sensitivity of the material they exchange and the risks they associate with features operating at the device level.

Implications beyond personal messaging

For businesses, the issue is relevant wherever employees use personal or managed devices to discuss confidential work. Messaging applications can protect communications within their own systems, but an organization’s broader security approach also depends on device settings, approved software and rules for handling sensitive information. Signal’s announcement does not establish that enterprise data has been exposed; it highlights a boundary that security teams may need to consider as operating systems add AI capabilities.

Organizations may want to review how their device policies address screen capture and AI tools, especially when staff use messaging apps for sensitive discussions. At the same time, Signal’s change cannot substitute for those policies or resolve the broader question of what access an operating system should have to app content. The announcement brings that trade-off into sharper view: users expect helpful device features, while privacy tools seek to keep sensitive information under their control.

For now, Signal’s response is a targeted adjustment to screenshots on mobile, alongside its existing Windows default—not a claim that it can block every form of AI-enabled access. The measure’s effectiveness will depend in part on the limits of the threat it is designed to address, a point the company itself acknowledges. The larger question raised by Whittaker is whether app-level privacy guarantees can remain meaningful when operating systems gain broader access to what users see and do.