Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseInsight · On RadarZDNet AI→ original

Signal Blocks Screenshots on iOS 27 and Android to Keep AI Agents From Reading Chats

Signal is disabling screenshots in its iOS 27 app, and on Android it blocks them when an AI assistant tries to capture the screen. Built-in agents from Google and Apple can read everything visible on screen, and a decrypted chat can end up in their local database or on a server. The encryption holds, but the message has already been read. Signal introduced a similar protection a year ago because of Microsoft Recall. For developers of private apps, the screen should now be treated as an insecure boundary.

AI-processed from ZDNet AI; edited by Hamidun News
Signal Blocks Screenshots on iOS 27 and Android to Keep AI Agents From Reading Chats
Source: ZDNet AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceZDNet AI ↗
Editor-in-Chief
Key metric #1The protection does not work on macOS and does not close access through the file system
Key metric #2The block can be turned off, but Signal explicitly advises against it
Key metric #3A week earlier, Google DeepMind's AI agents formally verified Signal's code

Signal, the messenger, is blocking screenshots in its iOS 27 app, and on Android it blocks the screen capture at the moment an AI assistant tries to take one. The reason is that AI agents built directly into the operating system can "see" the screen and use what they see as context for answering user requests. Imagine spending years building a safe, then placing a camera next to it that photographs everything you take out of it. The encryption remains intact, but the decrypted conversation on the screen has already landed in someone else's lens. For anyone building apps with private data, this is a warning: the problem is not limited to messengers.

What Signal Is Changing

Signal President Meredith Whittaker wrote in a blog post on Friday that the company will block screenshots in the iOS 27 version. On Android, the block will activate when an AI assistant tries to capture the screen. The targets are system features from Google and Apple: Gemini Intelligence and Onscreen Awareness in Siri AI. They monitor screen content and feed it into queries as context.

For an ordinary user, this is convenient: you can ask "what should I reply to this message?" without copying the text. But for an encrypted messenger, the picture is different. According to Whittaker, when sensitive data is captured this way, it is retained and becomes accessible to any local database or remote server that the OS-level agent uses at that moment. After that, Signal has no control, and the privacy of everyone in the conversation is at risk, not just the phone owner's.

This is the second such case. A year ago, Signal blocked screenshots because of Microsoft Recall, a feature that could capture screens showing chats on Windows 11 devices. Back then, a single feature from a single company posed the threat; now several major players are building agents. Unconfirmed reports say Samsung is working on its own agent, Galaxy Claw, and Apple is preparing a rival to Claude and ChatGPT.

Why the Protection Is Incomplete

Signal does not hide the fact that blocking screenshots does not solve the problem in full. Agents can reach app data through other routes, for example by taking content already stored on the device through the file system. On macOS, the new measures do not work yet.

There is a more troubling part. Whittaker said that Signal and other developers have no reliable tools to stop agents from undermining privacy at the app level. Even when Signal manages to detect and block AI access, Google, Apple, Microsoft and others provide no way to restrict it without serious loss for the core user experience. In effect, third-party developers are patching holes with makeshift tools, while platform owners set the rules.

The issue is not abstract. This week, a Reuters investigation, supported by Signal and the AI Now Institute, showed that Meta's Muse agent used permissions granted during initial setup in unexpected ways. It accessed apps and data the user never explicitly connected to it. This is an example of how broad rights granted once turn into access to anything.

The contrast is telling. A week earlier, researchers at Google DeepMind used AI agents to formally verify Signal's code, building on the messenger team's own work, and effectively proved that the encryption works. That is a very high bar for guarantees. But, as ZDNET notes, it loses its meaning if an agent simply reads messages off the screen: encryption mathematics is powerless against someone who reads the already decrypted text.

What Users and Developers Should Do

Users can disable the restriction in settings if they need screenshots. Signal strongly advises against it: this would potentially make the contents of your conversations, and those of your contacts, available to agents, and privacy promises would stop applying. An important detail: you are making that decision for yourself and for the people you message.

Developers should do three things. First, treat the screen as an untrusted boundary: anything displayed can be read by an OS agent. Second, check what data the app stores in locations accessible through the file system. Third, monitor whether third-party apps receive proper mechanisms from Apple, Google and Microsoft to opt out of agent access. Until those exist, Signal's approach remains the option: block screen capture and accept the loss of convenience.

For the industry, the conclusion is simple. The deeper AI agents are embedded in the OS, the more often apps will pay for it with their own functionality. The dispute over who decides what an agent may see is only beginning.

Story

Key Questions Answered

What Signal Is Changing?

Signal President Meredith Whittaker wrote in a blog post on Friday that the company will block screenshots in the iOS 27 version. On Android, the block will activate when an AI assistant tries to capture the screen. The targets are system features from Google and Apple: Gemini Intelligence and Onscreen Awareness in Siri AI. They monitor screen content and feed it into queries as context. For an ordinary user, this is convenient: you can ask "what should I reply to this message?" without copying the text.

Why the Protection Is Incomplete?

Signal does not hide the fact that blocking screenshots does not solve the problem in full. Agents can reach app data through other routes, for example by taking content already stored on the device through the file system. On macOS, the new measures do not work yet. There is a more troubling part. Whittaker said that Signal and other developers have no reliable tools to stop agents from undermining privacy at the app level.

What Users and Developers Should Do?

Users can disable the restriction in settings if they need screenshots. Signal strongly advises against it: this would potentially make the contents of your conversations, and those of your contacts, available to agents, and privacy promises would stop applying. An important detail: you are making that decision for yourself and for the people you message. Developers should do three things. First, treat the screen as an untrusted boundary: anything displayed can be read by an OS agent.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…