This is a MapMyTimes opinion piece. It responds to Signal's announcement on 20 November 2026 that it will stop built-in AI assistants from reading its users' chats, blocking screenshots inside the app on iOS 27 and switching on its screen-security setting on Android when it detects an AI assistant trying to read the screen. MapMyTimes reported the announcement the same day.

In a blog post by Signal president Meredith Whittaker, the messaging service named Google's Gemini Intelligence, Apple's Siri AI and its Onscreen Awareness feature, and Meta's Muse agent as the kinds of system-level and third-party agents it is guarding against. It admitted its protections are incomplete: blocking screenshots does not stop other routes, such as reading content already stored on the device, and on macOS Signal says it cannot reliably block screenshots at all, so it can only advise users to exclude Signal in their Siri settings.

Why does this matter to people who do not use Signal?

It matters because the problem Signal describes applies to every private app on a phone, from banking and health apps to the chats where families share documents. End-to-end encryption protects a message while it travels. Once the message is on the screen, an AI agent with permission to see the screen can see it too. Users have been told for years to look for encryption; the next question is who else is reading the screen.

The concern is not hypothetical. Reuters reported, as carried by TimesLIVE, that Meta's Muse agent read Signal messages on an Apple computer and an Android phone by using broad operating-system permissions, including accessibility and screenshot tools, and that Apple then released an update that effectively closed that particular pathway and added new permissions and warnings for such agents.

What should Apple, Google and Microsoft do?

Signal's requests are modest and, in our view, correct. It wants easy, well-documented ways for developers to opt their apps out of AI access, covering screen reading, files, notifications and other system features, without breaking the app for users. It also wants clear technical documentation of how AI agents access and process data, and tools to stop third-party agents from using accessibility permissions to reach app data. Signal says it had to investigate how these agents behave on its own because the documentation is lacking.

A standard opt-out would be better than each app inventing its own workaround. Signal's Android approach, for example, depends on detecting AI access because it cannot reliably detect screenshots on many manufacturers' devices. A banking app or a hospital app should not have to reverse-engineer an operating system to protect its customers.

Is there a fair case for the AI agents?

Yes. Meta told Reuters that agents need to see and interact with people's computers to work on their behalf, and that Muse reaches other apps only through permissions users explicitly grant. Many people genuinely want an assistant that can summarise a busy group chat or act on a message for them, and accessibility features exist precisely because some users depend on software that can read the screen. Blunt blocks risk breaking those uses.

That is why the fix should be precise rather than blanket. Signal itself lets users turn the protection off, while warning that doing so may expose their own and their contacts' messages to AI agents. A good system would let each user decide, app by app, what an assistant may see, and would let an app declare that its content is sensitive by default. Consent given once during setup, for broad permissions most people do not read, is not the same as informed choice.

What should users do meanwhile?

Users can check which AI assistants and accessibility services on their phones have permission to read the screen, and keep in mind that a chat is only as private as the least careful person in it. But the main responsibility lies with the companies that build the operating systems. Signal says it is cautiously optimistic that Google and Apple are engaging. They should turn that engagement into a published, standard opt-out that any app can use, before AI agents become a default feature on every phone.

Photo source