SIAgentNews.com
← All news
DevelopingAgentsSecurity

Signal blocks mobile screenshots to counter AI surveillance

The encrypted messaging app is restricting screen captures on Android and iOS to prevent OS-level AI agents from scanning private conversations.

· First reported by TechJuice

The short version

  • Signal now blocks screenshots on Android and iOS to prevent AI agents from scanning on-screen content.
  • The move follows a similar restriction on Microsoft's Recall feature on Windows.
  • On Android, users will see a blue privacy screen; on iOS, content is hidden in the app switcher.
  • Users can manually disable these protections, though Signal warns this breaks their privacy guarantees.
  • The company is calling for OS providers to give developers better tools to block AI-level data access.

Why it matters

This signals a shift in the privacy landscape where the threat is no longer just the app itself, but the operating system's own AI features. It highlights a growing conflict between convenient AI assistants and the fundamental right to private, unmonitored communication.

The threat of agentic AI

Signal has officially begun blocking screenshots on both Android and iOS devices. This change comes as a direct response to the rise of agentic AI systems being integrated into major operating systems like Windows, Android, and macOS. These AI agents are designed to be helpful assistants that can scan and interpret what is currently visible on a user's screen.

Signal president Meredith Whittaker argues that these capabilities pose an existential threat to private communication. Because these AI agents often have broad permissions, they can capture sensitive data and store it in local databases or send it to remote servers. Once the data is captured by the operating system, Signal loses the ability to control or protect that information.

The company describes this as a fundamental shift in the digital environment. They argue that independent developers can no longer make firm privacy promises when the underlying platform is designed to monitor application content. Whittaker recently described the current state of OS integration as a situation where the "water we swim in will have been poisoned."

How the blocks work

The new restrictions function differently depending on the mobile platform being used. On Android, the app utilizes a Screen Security feature. When a user attempts to take a screenshot while Signal is active, the system will display a blue privacy screen featuring the Signal logo.

The experience on iOS is handled through the app switcher. Signal has implemented a feature that hides the app's content when a user swipes up to view their open applications. This prevents the OS from taking a snapshot of the conversation for the multitasking preview. Both methods are intended to keep chat content invisible to the system's automated processes.

While these features are enabled by default, they are not absolute. Signal has provided a way for users to disable these restrictions within their privacy preferences. However, the company explicitly warns that turning these features off will break the privacy guarantees that the app normally provides to its users.

A pattern of resistance

This mobile update is the latest step in a broader campaign by Signal to defend against automated data scraping. In May 2025, the company began blocking Microsoft's Recall feature on Windows. Recall is a tool that takes screenshots of the user's screen every few seconds to feed into an AI model for later searching.

To stop Recall, Signal applied a Digital Rights Management (DRM) flag to its Windows application. This is a similar technology to what Netflix uses to prevent people from taking screenshots of movies.

Escalating warnings on AI risks

Meredith Whittaker has consistently raised alarms about the dangers of agentic AI throughout 2026. During a meeting at Davos in January, she warned Bloomberg that AI agents granted root access permissions are particularly perilous for secure applications. These permissions allow software to operate with high-level authority over a device's core functions.

Whittaker expanded on these concerns at the TechBBQ conference in Copenhagen. She used blunt terms to describe how AI integration into operating systems can compromise security. She specifically noted that asking an AI assistant to summarize screen content triggers a screenshot of the active app, which could include private Signal messages.

Industry responses and alliances

The push for better security has already prompted some reactions from major tech providers. For example, Apple tightened access controls in the macOS 27.3 beta 2 following reports that Meta's Muse AI assistant was able to read iMessages. This occurred after it was discovered that the assistant could access messages through Full Disk Access permissions.

Signal's latest blog post acknowledges that they are not alone in identifying these risks. The company expressed gratitude to its allies who are advocating for similar protections across the tech ecosystem. They are calling for a broader recognition of how current AI integrations represent an existential threat to privacy at the application layer.

What to watch

  • How other encrypted apps like WhatsApp or Telegram respond to OS-level AI scanning.
  • Whether Apple and Google release new developer tools to allow apps to opt out of AI agent scanning.
  • Further reports on how Meta's Muse AI assistant interacts with other encrypted platforms.

Try it

Check your Signal privacy settings on your mobile device to see if the screenshot and app switcher protections are enabled by default.

Sources and links

Read the original at TechJuice

First reported by TechJuice. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

More AI news