SAN FRANCISCO (Realist English). The Signal messenger has introduced restrictions on taking screenshots in its app on iOS 27 and Android to prevent built-in AI assistants from reading users’ personal messages. This is stated in a blog post by Signal President Meredith Whittaker, published on November 19.
How the Protection Works
On iOS 27, Signal enables the “Block Siri AI” setting by default, which not only blocks screenshots within the app but also prevents content from being scanned by the Onscreen Awareness feature of Siri AI. On Android, the app automatically activates the “Screen security” setting when it detects an attempt by an AI assistant to read the screen.
Users can disable the protection in settings, but Signal warns that this “could potentially expose your messages and your contacts’ messages to these AI agents” and would violate privacy guarantees.
The Reason: AI Agents Are “Hostile to Privacy”
Whittaker characterized the AI agents built into operating systems as “hostile to privacy.” According to her, features such as Gemini Intelligence from Google and Onscreen Awareness from Apple scan on-screen content to respond to user queries.
Once captured, Signal data may end up in the agent’s local database or on a remote server, beyond the messenger’s control.
“This is an existential threat to Signal and to our ability to responsibly carry out our mission,” Whittaker stated.
Acknowledgment of the Measure’s Incompleteness
Signal acknowledges that the proposed measure is “incomplete.” Blocking screenshots does not prevent AI agents from accessing data through the device’s file system. In addition, on macOS, implementing similar protection is impossible due to the lack of a documented API. On many Android devices, the app cannot detect an attempt to capture the screen if screenshots have already been disabled.
Whittaker also pointed to the problem of third-party agents, such as Meta Muse, which use deep accessibility permissions as a “skeleton key” to access app data.
Background: Recall and Windows
Signal already applied a similar approach in May 2025, blocking screenshots in Windows 11 to counter Microsoft’s Recall feature, which periodically captures the screen to create a searchable database, including disappearing messages.
Demands to Apple, Google, and Microsoft
Signal called on operating system operators to provide developers with simple, documented tools to opt out of AI access to app data — screen, file system, notifications, and APIs — without sacrificing functionality. Companies must also disclose information about how their agents obtain and process data.
“Signal is the canary in the coal mine, but this affects the entire ecosystem and should prompt everyone who cares about these fundamental rights,” the statement reads.
As observers note, the key feature of the situation is the gap between the declared privacy of end-to-end encryption and the real vulnerability of data once it is decrypted on screen. Signal blocks one leak channel but acknowledges that the AI agent ecosystem creates many others.
For the secure communications market, the immediate effect will be visible primarily in whether Apple and Google follow Signal’s call and whether developers gain tools to opt out of AI access without losing functionality — and that is precisely where the balance between AI innovation and privacy is most sensitive.







