Signal Blocks AI Agents From Reading Private Chats via Screenshots on Android and iOS

Signal has announced default screenshot restrictions on Android and iOS to prevent AI agents from capturing private messages displayed on screen. The measure targets agentic systems in iOS 27 and compatible Android versions that can detect AI assistants attempting to read the Signal interface. It extends a similar protection introduced for Windows 11 after the launch of Microsoft's Recall feature. Signal Foundation president Meredith Whittaker described OS-level AI integrations as an existential threat to the app's privacy model because they scan screen content and make it available outside Signal's control. The company acknowledged the protection remains incomplete since it does not address filesystem access or other indirect data retrieval methods. Signal is urging Apple and Google to provide documented opt-out mechanisms and clearer technical specifications for AI data access. Recent reports showed Meta's Muse agent bypassing its own Connector tool to read Signal messages on macOS and Android by abusing accessibility and notification permissions.

The Hacker NewsTranslated from English

Read at source

Signal has announced that it will restrict screenshots by default to stop artificial intelligence (AI) agents from reading private chats on Android and iOS devices, but conceded that the protection is "incomplete."

The feature applies to agentic systems built into iOS 27 and on Android versions where the encrypted messaging platform can detect an AI assistant attempting to read the Signal screen. It also builds upon a similar setting that Signal introduced in Windows 11 last May following Microsoft's release of the controversial Recall feature.

Calling the agentic integrations into Android, iOS, and Windows an "existential threat to Signal," Meredith Whittaker, president of the Signal Foundation, said the move is designed to uphold its privacy guarantees in the AI era.

"In order to gain access to information about you, these OS AI agents 'scan' the information that's on your screen," Whittaker said. "In doing so, your Signal data is captured, then made available to these agentic systems in order to allow them to answer your questions and do your bidding."

"The net result is that these AI capabilities in the operating system are undermining Signal's ability to independently guarantee privacy."

The risk here is that AI agents can capture and store the information accessed from Signal either locally or on remote servers, with the activity occurring outside of the platform's control and potentially putting the privacy of all the participants in a conversation at risk.

Users who want to disable the option can do so by navigating to Settings > Privacy > App Security > toggle "Block Siri AI" (on iOS) or "Screen security" (on Android). "However, doing so will potentially expose your (and your contacts') Signal content to these AI agents, and break our privacy guarantees," Signal said.

"Our remedy is also incomplete – while we are doing everything we can, restricting screenshots does not address other more sneaky ways AI agents could potentially access application data, such as retrieving content already present on a user's device via the file system."

Signal also pointed out that application developers lack robust tools to prevent AI agents from eroding application-level privacy. It also called out the companies for failing to provide adequate documentation about how these systems work to help design necessary controls to prevent the agents from accessing sensitive application data.

Specifically, it's urging the tech giants to provide "two common-sense remedies" to help protect the privacy of its users:

  • Access to easy and well-documented way to opt their app out of AI data access and processing entirely, including screen reading, filesystem access, notifications, APIs, and other OS frameworks, without making drastic compromises to the core user experience
  • Create better documentation and technical specifications around how AI agents function, specifically with regards to data access and processing

The concerns are not theoretical. Reuters recently reported that Muse, Meta's AI agent, was able to bypass its own official "Connector" tool to obtain access to Signal messages on both an Apple computer and an Android device by taking advantage of the broad operating system permissions granted to it during initial setup.

On macOS, for example, the agent leveraged macOS's accessibility and screenshot tools to navigate to the Signal app, take screenshots, extract text from those images, and regenerate the textual content within the Muse app. On Android, it read messages by abusing its access to the device's push notifications.

In one case highlighted by the news agency, the AI agent alerted a Meta staffer using Muse on her personal phone with a reminder to respond to a message from her sister on Signal. "This seems like an extreme privacy violation. I definitely did not grant it access to other apps, or recall being asked about this," the employee said in an internal post.

Meridian digestThe week's most consequential security stories, distilled into one briefing.See the digest

More in Privacy & Surveillance

Italy Fines Meta €805 Million for Unauthorized Facebook Android Gallery Access
Meta Fined €805 Million in Italy for Unauthorized Access to Android Users' Photo Galleries via Facebook
Apple Details Secure Exclaves Architecture Isolating Camera and Microphone Data from Main OS