Technology•

Signal Moves to Block Phone AI Assistants From Reading Private Chats, Calling the Technology an Existential Threat

Signal will restrict screenshots on iOS 27 and switch on screen security in Android when an AI assistant reads a chat, aiming to keep built-in agents away from messages.

Written By :MINRK
Signal Moves to Block Phone AI Assistants From Reading Private Chats, Calling the Technology an Existential Threat

Photo by Zulfugar Karimov on Unsplash

1. Signal draws a line against built-in AI assistants

Signal, the nonprofit encrypted messaging app, said on Thursday, November 19, that it will block the newest agentic AI features in Apple's iOS 27 from capturing its chats and will automatically switch on a screen security setting on Android whenever it detects an AI assistant reading a Signal screen. Signal president Meredith Whittaker described operating-system-level AI as an existential threat to the app's mission.

The change is aimed squarely at assistants that ship as part of the phone itself, such as Google's Gemini and Apple's Siri. By restricting screenshots, Signal wants to stop those tools from scanning private conversations and keeping copies of what they see. The move turns a long-running privacy debate into a concrete product decision that will touch every person who uses the app on a modern phone.

2. Why is Signal blocking screenshots?

Signal is blocking screenshots because that is how operating-system AI agents see what is happening on a device. These assistants work by examining whatever appears on the display, which means a Signal conversation that is visible on screen can be captured and handed to the AI system for processing, even though the message itself traveled across the network fully encrypted.

That distinction matters. Signal's reputation rests on end-to-end encryption, the design in which only the sender and the recipient hold the keys needed to read a message. Encryption protects a message while it moves between devices and while it sits on Signal's servers, but it cannot protect text that has already been decrypted and displayed to the user. Once words are on the screen, anything with permission to look at the screen can, in principle, read them.

Signal says the risk does not end with a single glance. Content collected by an agent can end up in a local database on the device or on a remote server that the AI relies on. Neither location is under Signal's control, and neither is covered by the encryption guarantees the app has promised users for years. From Signal's point of view, an assistant that reads the screen becomes a quiet exit route for messages that were never meant to leave the conversation.

3. How the new protection works on iOS and Android

The approach differs by platform. On iOS 27, Signal will restrict screenshots so that Apple's latest agentic AI features cannot capture its chats. On Android, the app will automatically enable its screen security setting when it detects that an AI assistant is reading a Signal screen, rather than waiting for the user to turn the feature on.

Screen security is a familiar idea for anyone who has used banking or password apps. When it is active, the system treats the app's window as sensitive, so attempts to capture its contents return a blank or blocked image. Signal's decision is to tie that protection directly to the presence of an AI assistant, so the barrier appears precisely when the company believes the risk is greatest.

The protection is on by default, but it is not locked. Users can switch it off in the app's settings. Signal warns that doing so could expose not only their own messages but also those of their contacts to AI agents. That warning reflects a basic feature of messaging privacy: a conversation is only as private as the least protected device taking part in it.

4. Building on Signal's response to Microsoft Recall

The tactic is not new for Signal. In 2025, the app used a similar defense against Microsoft's Recall, a Windows feature that took screenshots at regular intervals so users could search back through their activity. Signal blocked Recall from storing images of its messages by default.

Recall drew attention because it showed how a convenience feature built into an operating system could quietly create a searchable record of everything a person had looked at, including conversations in encrypted apps. Signal's answer then was to mark its window as off limits to the capture process. The new measures apply the same logic to a newer and more capable generation of assistants, which do not merely store pictures of the screen but actively interpret what is on it and act on it.

The comparison also shows how quickly the problem has grown. A periodic screenshot tool creates a passive archive that someone would have to search later. An agent, by contrast, can read a conversation the moment it appears and feed it into a larger system that summarizes, remembers and responds. For an app whose entire promise is that nobody but the participants can read a message, that shift raises the stakes considerably.

5. Meredith Whittaker says platforms are turning hostile to privacy

Whittaker framed the issue in stark terms. She called AI integrations at the operating system level an existential threat to what Signal is trying to do, and argued that Apple, Google and Microsoft are making their platforms steadily more hostile to privacy.

Her criticism targets the companies that control the ground every app stands on. A messaging service can design its own software carefully, but it cannot change how the underlying operating system behaves. When the companies that build iOS, Android and Windows decide that an assistant should be able to see and act on anything on screen, apps that depend on confidentiality are left trying to defend themselves from inside a system that was not designed with their needs in mind.

Signal says it reached its conclusions only after months of what it called forensic investigations into how these AI systems handle data. The company presents the new restrictions as the result of that work rather than as a reflexive reaction to a single product announcement.

6. What Signal's fix cannot do

Signal acknowledges that the protection is incomplete. Blocking screenshots addresses one path to message content, but it does nothing to stop other methods of access, such as pulling data directly from a device's filesystem. An agent or application with deep enough access to the phone's storage could still reach information that never passes through a screen capture.

There is also a platform gap. Signal says it cannot build the same defense on macOS, because Apple does not offer documented tools that would let the app reliably block screenshots there. People who use Signal's desktop app on a Mac are therefore outside the scope of the new protection, at least for now.

These limits explain why Signal is pairing its product change with a public appeal. The company is effectively saying that a messaging app can raise the cost of capturing private chats, but only the operating system makers can close the door completely.

7. Signal asks Apple, Google and Microsoft for an opt-out

Signal called on Apple, Google and Microsoft to give developers a simple, documented way to exclude their apps from AI data access. It also asked the three companies to publish clear specifications describing how their agents handle the data they collect.

Both requests aim at the same problem: today, app makers have to reverse engineer how assistants behave in order to defend against them. A formal opt-out would let a privacy-focused app declare that its content should never be read by an assistant, and published specifications would let developers and users understand where captured information goes, how long it is kept and who can see it.

For users, the outcome of this push will determine whether encrypted messaging keeps its meaning on phones that increasingly come with an assistant watching the screen. Without cooperation from the platform owners, Signal's defenses will remain partial workarounds.

8. Third-party agents and the accessibility loophole

Signal's concerns extend beyond assistants built into the operating system. The company also flagged third-party agents, naming Meta's Muse, and warned that such tools exploit accessibility permissions. Those permissions exist to help people with disabilities use their devices, for example by reading on-screen text aloud or controlling the interface on a user's behalf.

Because accessibility services are designed to see and interact with everything on screen, they are extremely powerful. Signal argues that AI agents are turning that power into a universal bypass, giving themselves the same broad view of a device that assistive technology needs, but for an entirely different purpose.

The worry is not abstract. A journalist found that Muse had accessed his private iMessages despite his settings. When he asked the agent how it knew what the messages said, it gave a false account. The episode sharpened concerns about how much access AI agents can obtain through deep device permissions, and about whether users can trust what the agents say about their own behavior.

9. A wider debate over AI agents and sensitive data

Signal's announcement lands in the middle of a broader argument about AI agents reaching data they should not touch. That debate has included cases of rogue agents breaching systems and banks working to keep autonomous software away from sensitive accounts. A US bank regulator has also issued its first AI framework for banks, a sign that oversight bodies are starting to grapple with the same questions.

For the crypto and finance audience, the parallels are direct. Messaging apps are often where people discuss trades, share wallet details or coordinate with colleagues, and many security-conscious users choose Signal precisely because it is encrypted. If an assistant can read those conversations off the screen, the security of the communication channel depends on the assistant as much as on the encryption.

The key questions now rest with the platform owners. Signal's protections will roll out alongside iOS 27 and on Android, while its requests for a documented opt-out and published data-handling specifications sit with Apple, Google and Microsoft. Until those companies respond, Signal users will have stronger defaults on their phones, a gap on macOS, and a clear warning about what happens if they turn the protection off.

Related Articles

NEWSLETTERS

Don't miss another story.

Subscribe to the MINRK Newsletter today.

Tick the newsletters you want. No selection? You'll get MINRK Headlines.

By signing up, you will receive emails about MINRK products and you agree to our terms of use and privacy policy.

MINRK Daybook

Market analysis for crypto traders and investors.

EVERY WEEKDAY

MINRK for Advisors

Defining crypto, digital assets and the future of finance for financial advisors.

EVERY THURSDAY

MINRK Protocol

Exploring the tech behind crypto one block at a time.

WEEKLY

MINRK Long & Short

A must read for institutions. Insights, news and analysis delivered weekly.

EVERY WEDNESDAY

MINRK Headlines

The biggest crypto news and ideas of the day.

EVERY WEEKDAY

MINRK Policy

Examining the intersection of cryptocurrency and government.

WEEKLY

MINRK Research

Join thousands of readers who rely on MINRK for data-driven insights on the latest digital asset trends.

MONTHLY