Signal blocks Siri, Gemini and other AI agents from reading chats
The encrypted messaging app now stops iOS and Android system assistants from capturing on-screen messages, extending a similar move it made on Windows against Microsoft Recall.

Key takeaways
- Signal now blocks screenshots in the app by default on iOS 27 and stops Android AI assistants from capturing its screen, with an opt-out in settings.
- Signal says OS-level agents such as Gemini and Siri AI can capture open chats and send the content to local databases or remote servers beyond its control.
- The move extends Signal's earlier Windows decision, where screenshot blocking was enabled by default to counter Microsoft's Recall feature.
- Signal admits the fix is incomplete, offers no macOS protection yet, and is asking Apple, Google and Microsoft for better developer tools.
In this article
Signal blocks AI agents on iOS and Android

Signal has begun blocking the built-in AI assistants on iOS and Android from reading its users' messages, according to Ghacks. The change arrives through updates to the encrypted messaging app and follows a similar move the company made on Windows last year.
On iPhones running iOS 27, screenshots are now blocked inside the Signal app by default. On Android, the app stops screen capture when it detects an AI assistant attempting to read what is displayed. Users who want to take screenshots can switch the protection off in Signal's settings, but the app warns that this could expose their own messages and those of their contacts to AI agents.
Why Signal sees operating system assistants as a threat
Signal argues that Google and Apple are building agentic features, meaning software that can act on a user's behalf, deep into their operating systems. Examples include Google Gemini with its Gemini Intelligence features and Siri AI with Onscreen Awareness on Apple devices.
These assistants work by examining what appears on the display in order to answer questions or carry out tasks. If a Signal conversation is open at that moment, its content can be captured and passed to the assistant. From there, Signal says, the data may end up in a local database on the device or on a remote server, and the company has no way to control what happens next.
In a blog post announcing the change, Signal described the situation as an existential threat to its ability to guarantee privacy. Its encryption protects messages while they travel between devices, and its code is open source, but neither of these helps if an operating system feature simply reads the decrypted content straight off the screen.
The Windows precedent with Microsoft Recall
This is not the first time Signal has taken this kind of step. Last year, the app turned on screenshot blocking by default on Windows to stop Microsoft's Recall feature from capturing and storing images of Signal conversations on Windows 11.
Recall is a Microsoft feature that regularly takes screenshots of a user's activity so it can be searched later. Signal's earlier decision on Windows set the pattern for the new iOS and Android protections: prevent the capture by default, and let users opt out if they accept the risk.
Gaps in the protection
Signal acknowledges that the current fix is incomplete. Blocking screenshots does not cover every route an AI agent could take. For example, an agent might reach app data through the file system rather than the screen.
There is also no equivalent protection on macOS yet. Signal says the technical tools needed to build it do not currently exist without breaking the app. On top of that, third-party assistants such as Meta's Muse can use accessibility permissions, which are meant to help assistive technology, to read the contents of other apps. Signal wants Google and Apple to defend against this kind of access as well.
Signal calls on Apple, Google and Microsoft
Signal describes it as a serious problem that app developers lack reliable tools to stop AI agents from undermining privacy at the application level. In its view, Google, Apple and Microsoft give developers no straightforward way to restrict what AI features can access without making drastic compromises to the core user experience.
According to Signal, Google and Apple are already in talks with it about how these systems should interact with privacy-focused apps, and it hopes to share more soon.
For now, users on iOS 27 and Android get the new protection automatically, with an opt-out in settings. Those on macOS remain without a technical safeguard until the platform offers a workable one.
Sources
Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.



