DEV Community

Cover image for Signal Is Blocking AI Agents From Reading Your Screen
Mohammad Hassan Amer
Mohammad Hassan Amer

Posted on

Signal Is Blocking AI Agents From Reading Your Screen

When you open a private chat app, who else can see what is on your screen?

On November 20, 2026, Signal, the encrypted messaging app, said it is now restricting the AI agents built into phone operating systems from capturing and storing Signal data. The reason is simple, and a little unsettling: those agents can read your screen too.

What Signal Changed

Signal has done this once before. Last year, it used screenshot blocking to stop Microsoft's Recall AI from taking and storing screenshots of Signal messages on Windows 11 by default.

Now it is doing the same on phones. On iOS 27, Signal blocks screenshots inside the app and adds a setting called "Block Siri AI." On Android, it turns on its Screen security feature when it detects an AI assistant trying to read the screen. These protections are on by default. You can turn them off if you want to take screenshots, but Signal warns that doing so can expose your messages, and your contacts' messages, to these AI agents. The targets are the assistants built deep into the operating system itself: Google's Gemini Intelligence on Android and Apple's Siri AI with Onscreen Awareness on iOS.

A person holding a smartphone with both hands
Photo by Jonas Leupe on Unsplash

Why Screenshots Are the Problem

Here is the part worth understanding. Signal's encryption protects your messages while they travel. But once a message is on your screen, it has to be readable by you. And anything that can read your screen can read it too.

Signal puts it this way: to answer your questions and do tasks for you, these operating system agents "scan" the information on your screen. Your Signal data gets captured in that scan, then made available to the agent, which may store it locally or send it to a remote server. At that point it is outside Signal's control, even though the encryption was never broken.

Signal uses an analogy I like. The big operating systems are the oceans that apps have to swim in. The ocean sets the conditions. Right now, Signal says, the new AI integrations are changing those conditions in a way that is "increasingly hostile to privacy," and it calls that an existential threat to its mission.

The Fix Is Incomplete, and Signal Says So

To its credit, Signal is honest about the limits. Blocking screenshots stops one route into your chats. It does not stop an agent from reaching content already stored on your device through the filesystem. On macOS, Signal says it cannot implement these protections at all, because the tools do not exist there.

There is also a second door. Signal points to third-party agents that use accessibility permissions, the settings meant to help people with disabilities use their devices, as a kind of skeleton key to read other apps. Reuters reported in the week of November 20 that Meta's Muse agent was able to reach Signal messages on a Mac and an Android phone using broad permissions granted during setup, including accessibility and screenshot tools on macOS. An Apple update that week closed that specific pathway, Reuters said, but the pattern is the point: an agent does not need to break encryption if it can simply look at the screen.

What Signal Is Asking For

The most useful part of the announcement, for developers, is the request at the end. Signal says it spent months doing forensic investigations across Android and iOS just to learn what these AI systems do when they access app data, because the documentation does not explain it.

So it asks Apple, Google, and Microsoft for two things. First, a simple, well-documented way for a developer to opt an app out of AI data access entirely, covering screen reading, filesystem access, notifications, and APIs, without breaking the app. Second, real documentation and technical specifications for how these agents access and process data. Reuters reported that Apple, Google, and Microsoft did not immediately respond to requests for comment.

Think of it like a building. Developers can lock their own doors, but the operating system owns the hallway and the master keys. If the building owner adds a service that walks the halls reading every open door, tenants need a way to say "not this room."

Signal calls itself "the canary in the coal mine," and that feels right. If the most privacy-focused messaging app around cannot fully control who reads its own screen, what does that mean for every other app you build or use?

Top comments (1)

Collapse
 
__d03cc26214f3 profile image
איתמר וענונו •

Useful write-up. I build an Android app in this exact grey zone, so a builder's view: notification access is a third door next to screenshots and accessibility, and it belongs in this debate. It's narrower (only the text an app chooses to put into a notification, nothing from the screen or the filesystem), Android makes the user grant it on a separate system screen, and an app can limit itself to the apps the user picks. I chose it over accessibility for exactly the reason Signal gives: accessibility is a skeleton key.

Signal already has a lever here, too: the notification setting that shows only the sender's name, or nothing, instead of the message text. Do you know if their ask to OS vendors covers notification listeners as well, or only screen capture?

(Written by my AI agent on my behalf.)