Signal Blocks Screenshots to Stop AI Agents Reading Chats

Signal has started blocking screenshots inside its messaging app by default, in a direct response to the AI assistants now built into phones that can read whatever appears on screen. The encrypted messaging service announced the change on 20 November 2026, warning that the new generation of operating-system AI agents poses a serious threat to private conversation.
Signal president Meredith Whittaker set out the move in a blog post, saying the company could no longer assume that what users see on their phones stays on their phones. Google and Apple have woven agentic AI deep into their mobile operating systems, and those assistants work by scanning the contents of the screen.
What changes on iPhone and Android
On iPhones running iOS 27, Signal will now block screenshots inside the app by default. It is also adding a setting called Block Siri AI, designed to stop Apple's assistant from reading message content on screen.
On Android, Signal takes a different approach. When it detects an AI assistant trying to read the screen, it automatically switches on its existing Screen Security feature. Signal acknowledged that this detection does not work on every version of Android, so the protection will not reach all users at once.
Users who still want to take screenshots of their chats can switch the protection off in Signal's settings. Signal warns against doing so, saying it could expose their own messages, and messages sent by their contacts, to AI assistants.
Why Signal acted now
The concern is how these assistants gather information. To answer questions and carry out tasks, they capture what is displayed on screen, often using optical character recognition to extract the text. That means a private Signal conversation can be read, stored and sent to a remote server, even though Signal's own encryption keeps it hidden in transit.
Whittaker wrote that Google and Apple are integrating agentic capabilities such as Gemini Intelligence and Siri's Onscreen Awareness deep into their operating systems. Once a message has been scanned by one of these systems, Signal can no longer guarantee it stays private.
"We want to be very clear: it is a serious problem that Signal — and application developers beyond — lack robust tools to prevent AI agents from undermining application-level privacy," Whittaker said.
A partial fix, by Signal's own admission
Signal is open about the limits of what it can do. Blocking screenshots closes one route into a conversation, but it does not stop an AI agent reaching data already stored on the device through the file system. It also cannot help if the person at the other end of a chat has switched the protection off, allowing an assistant on their phone to read the same messages.
Third-party agents raise a further problem. Some arrive with broad permissions granted during setup, including accessibility access that lets them interact with other apps. Reporting on the announcement named Meta's Muse agent as an example of that wider risk.
On Mac computers, Signal says it has no reliable way to block screenshots at all. Its advice for Mac users is to add Signal to the excluded-apps list under System Settings, then Siri, so the assistant cannot proactively read message content.
Not the first time Signal has done this
The move follows a similar step Signal took last year on Windows 11, when it blocked screenshots to stop Microsoft's Recall feature from capturing its window. Recall periodically records images of what a user is doing, and Signal acted to keep its chats out of that record.
Signal is now calling on Apple, Google and Microsoft to give developers proper tools to opt their apps out of AI data access entirely, covering screen reading, notifications, file-system access and system programming interfaces. Until those tools exist, it says, app makers are left patching around systems they cannot control.
The change lands as AI agents spread further across phones and computers. Other companies are drawing their own lines: Amazon has blocked some shopping agents from its store, United Airlines has banned transactions from unauthorised AI bots, and Anthropic has restricted third-party software from using its subscription accounts. Signal's step is one of the clearest yet from a messaging app, and it puts the tension between helpful assistants and private conversation in plain view.