Signal restricts screenshots to protect messages from AI agents
The messaging service limits screen access on iPhones and some Android devices, but acknowledges that the protection is not complete.
Listen to this article
There is no audio version yet. Request one and an AI voice will read the article aloud.
Read by an AI voice.
Signal will block screenshots in its mobile app when AI assistants try to read the contents of the screen. Signal says the measure is intended to prevent messages from leaving the encrypted app and reaching operating systems or agents acting on users’ behalf.
Signal announced the change on Friday. On iOS 27, screenshots in the app will be blocked by default. On Android, Signal tries to detect when an AI assistant wants to read the screen and activates screen protection for that purpose.
Users can disable the block. This allows them to take screenshots, but Signal says it also exposes messages to AI systems that can capture screen content. According to Signal, that choice could therefore also affect the privacy of conversation partners.
Signal points to the growth of so-called AI agents as the reason, noting that they do more than provide answers: they also perform tasks and combine information from multiple apps. According to Signal, such systems can use screens, notifications, files and accessibility features to extract messages from a secure environment.
Reuters previously reported that Meta’s Muse agent could gain access to Signal messages on an Apple computer and an Android phone through broad system permissions. Meta says such access is possible only after users have granted permission. This raises the question of how specific permission to access private conversations should be.
The protection has limits. Signal says it cannot reliably detect all Android versions and cannot introduce the block on macOS. The app also cannot prevent an agent from accessing data already stored on the device by other means. Signal is therefore asking Apple, Google and Microsoft for clear technical opt-outs and documentation.
For users, the main change concerns how the app works: screenshots may be unexpectedly refused, or a setting may need to be adjusted. Signal is also drawing attention to app makers’ dependence on the screen and access permissions provided by operating systems.
One story, several perspectives
What is established
- Signal is introducing screenshot and screen restrictions to counter AI agents.
- The protection does not work on all devices and can be disabled by users.
- Meta says agents gain access only through permissions granted by users.
Left
Arguments Control over screen data should not lie solely with large platform companies; users and independent apps should retain strong legal and technical privacy rights.
Values Privacy, protection of citizens, and the balance of power between platforms and users.
Consequences Without opt-outs, sensitive conversations could enter broader AI data streams; stricter rules could slow innovation in AI agents.
Centre
Arguments AI agents can provide useful accessibility and automation features, but access to private messages must be clear, limited and controllable.
Values Proportionality, transparency and workable technical standards.
Consequences Standardised permissions could combine security and ease of use, but would require time and cooperation between app and operating-system makers.
Right
Arguments Users should decide for themselves which assistants they grant access to, and companies should have room to build new services without a dense regulatory web.
Values Individual choice, innovation and limited government intervention.
Consequences Greater choice could lead to faster development, but users would also bear more responsibility for the consequences of the access rights they grant.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The technical changes, limitations and context were checked directly with Signal and Reuters. Claims about access by AI agents are attributed to the companies involved and are not presented as an established general practice.
- confirmed Signal blocks screenshots on iOS 27 and detects certain AI screen access on Android. — Signal describes both measures in its blog. source
- confirmed Users can disable the protection. — Signal says users can switch off screenshot protection in the settings. source
- confirmed Signal cannot reliably implement this protection on macOS. — Signal explicitly identifies macOS as a limitation. source
- confirmed Meta’s Muse gained access to Signal messages through broad system permissions. — Reuters reports this and includes Meta’s response that the permissions were explicitly granted. source
Editor's note
Signal is the primary source for the new settings and identifies the limitations itself. Reuters confirms the context involving AI agents and gives Meta’s response; the technical risks have not been independently measured in all cases.Sources
- Signal is Serious About Protecting Your Privacy — Signal
- Exclusive-Signal restricts screenshots to block AI agents from reading messages — Reuters via Investing.com
More on this in Dutch media
- de Volkskrant — „signal privacy”
- NOS — „signal privacy”
- Het Parool — „signal privacy”