Signal is taking an unusual step to protect private conversations from increasingly capable AI assistants built into mobile operating systems: blocking screenshots by default. The encrypted messaging app says these AI agents can access information displayed on users’ screens, potentially exposing sensitive conversations to systems outside Signal’s control.
In a blog post published on November 20, 2026, Signal president Meredith Whittaker described the growing integration of AI agents into operating systems as increasingly hostile to privacy. She argues that while Signal can protect messages through end-to-end encryption, it cannot independently guarantee privacy if the operating system itself can capture and process what appears on a user’s screen.
The change specifically targets AI capabilities such as Google‘s Gemini Intelligence and Apple‘s Onscreen Awareness with Siri AI, which are designed to understand on-screen information and help users complete tasks using contextual information.
Signal is blocking AI agents from reading conversations
Signal is introducing screenshot restrictions on iOS 27 and Android devices to prevent operating system-level AI assistants from capturing message content.
On iOS 27, screenshot blocking is enabled within Signal to prevent screenshots from being taken. On Android, the approach is slightly different. Signal detects when an AI assistant attempts to access screen content and automatically enables its Screen Security setting when that activity is detected.

The distinction matters because Android comes in numerous manufacturer-specific versions, and Signal says it cannot reliably detect every screenshot attempt across all Android distributions. Instead of disabling screenshots across the board and potentially confusing users, it focuses on detecting AI access.
Users who still want to take screenshots can disable the protection through Signal’s settings. However, the company warns that doing so could expose their conversations, as well as their contacts’ messages, to AI systems accessing on-screen information.
This isn’t entirely new territory for Signal. The company previously implemented screenshot-blocking protections against Microsoft’s Recall feature on Windows 11, preventing the AI-powered feature from capturing and storing screenshots of Signal conversations.
End-to-end encryption isn’t the whole privacy story anymore
Signal’s concerns highlight an increasingly complicated problem with AI-powered operating systems.
End-to-end encryption protects messages while they travel between participants, ensuring that the messaging service itself cannot read their contents. But that protection doesn’t necessarily prevent an operating system from accessing information after it has been displayed on a device’s screen.
For example, an AI assistant that can inspect the screen to understand a user’s request might also encounter private messages, contact information, or other sensitive details displayed inside an encrypted messaging application.
According to Whittaker, this information can potentially be captured, stored, and made available to whatever local database or remote server an operating system-level AI agent uses. Once that happens, Signal has limited control over how the information is subsequently handled.
Signal also acknowledges that screenshot blocking is not a complete solution. AI agents could potentially access application information through other methods, including retrieving data already present in a device’s filesystem.
The company is also concerned about third-party agents such as Meta‘s Muse, which can use extensive accessibility permissions to interact with applications. These permissions were originally designed to make devices and applications more accessible to people, but Signal argues that they could also become another route for AI agents to access sensitive information.
Signal wants Apple, Google, and Microsoft to provide better controls
Beyond restricting screenshots, Signal is calling on the companies behind major operating systems to give application developers more meaningful control over AI access.
Whittaker says Apple, Google, and Microsoft currently lack sufficiently clear technical tools and documentation that would allow developers to prevent AI agents from accessing sensitive application data without compromising the normal functionality of their apps.
Signal is asking these companies to provide two major improvements.
First, developers should have a straightforward, well-documented way to prevent their applications from being accessed or processed by AI agents. That protection should cover more than screenshots, including screen reading, filesystem access, notifications, APIs, and other operating system frameworks.
Second, operating system developers should publish clearer technical documentation explaining how their AI agents access, capture, and process application information.
There is also a significant limitation on Apple’s desktop platform. Signal says it cannot currently implement the same protections against Siri AI on macOS because Apple does not provide reliable APIs for blocking the types of screenshots used by these agents.
For now, Signal recommends that macOS 27 users add the app to the Excluded Apps section under System Settings > Siri to prevent proactive access to message content by Siri AI.
Despite these limitations, Signal says it is engaging with Apple and Google about improving how operating system-level AI interacts with privacy-sensitive applications.
The bigger question is who controls your data
The tension between AI convenience and application-level privacy is becoming harder to ignore. Operating system-level assistants are increasingly designed to understand what users are doing across applications, making it easier to summarize information, answer questions, and coordinate tasks.
But that convenience also introduces a new layer of access that individual application developers may not be able to fully control.
Signal’s decision is particularly interesting because it challenges the assumption that end-to-end encryption alone is enough to protect private conversations in an AI-driven computing environment. Protecting the communication channel is one thing; controlling what happens when an operating system can inspect the displayed content is another.
Whittaker describes the situation as an existential threat to privacy at the application layer, arguing that independent developers need the ability to protect their users even when operating system providers introduce new AI capabilities.
Signal’s screenshot restrictions are an imperfect workaround, not a permanent solution. But they raise an important question for the entire industry: if an AI assistant can access information across your applications to make your device more useful, should individual apps have the right to keep that information entirely off-limits?
Discover more from GadgetBond
Subscribe to get the latest posts sent to your email.
