Signal Blocks Siri and Gemini Screen Access | Zeniteq
Loading
Zeniteq is now available on mobile for iPhone, iPad and Apple Watch.Download nowLearn more
Signal Blocks Siri and Gemini Screen Access
Signal’s mobile defenses restrict AI assistants from capturing private chats, but Android detection limits and a manual macOS exclusion leave users with settings to check.
Signal is restricting AI-assistant screen access on mobile devices to prevent systems like Apple’s Siri and Google’s Gemini from harvesting private conversations.
By blocking screenshot capabilities on iOS 27 and implementing detection-based security on Android, Signal aims to stop AI agents from scraping displayed messages.
However, these defenses are incomplete, as they do not address filesystem access or other data interfaces.
Users should keep mobile protections enabled and manually add Signal to the excluded apps list in macOS system settings to ensure their sensitive communications remain outside the reach of automated agents.
Signal is restricting screenshots and AI-assistant screen access on mobile devices to keep private messages out of operating-system AI systems. In a November 20, 2026 announcement, Signal president Meredith Whittaker called the platform shift “an existential threat to Signal,” arguing that agents can move sensitive content beyond the messaging app’s control. (Signal)
The protections target Apple’s Siri AI on iOS 27 and Google’s Gemini and other assistants on supported Android configurations. Signal says screenshot blocking can prevent these systems from capturing conversations displayed inside the app, extending an approach it previously used against Microsoft’s Recall on Windows 11.
For readers, the immediate action is straightforward: keep the iPhone protection enabled, enable or retain Screen Security on Android, and exclude Signal from Siri on macOS 27. Those settings reduce exposure, though Signal explicitly acknowledges that screen blocking cannot cover every route an AI agent might use.
What Signal Users Should Enable on Each Platform
Signal on iOS 27 prevents screenshots inside the app, while Signal on supported Android configurations applies protections when it can detect an AI assistant attempting to read the screen. Users can disable screenshot blocking in Signal’s settings, restoring screenshots while potentially reopening that capture route. (Signal)
The platform differences matter. An iPhone default, an Android defense that depends on detecting access, and a user-managed macOS exclusion provide different levels of control.
Platform
Signal’s protection
What readers should do
iPhone running iOS 27
Screenshot blocking restricts Siri AI’s access to displayed Signal content.
Leave the default Block Siri AI protection enabled.
Supported Android configurations
Signal can block AI-assistant screen capture where it detects the attempt; behavior varies by device vendor.
Enable or retain Screen Security in Signal.
Mac running macOS 27
Signal says it cannot reliably implement equivalent protection itself.
Open System Settings → Siri → Excluded Apps and add Signal.
Signal’s announcement explains the mobile screenshot restrictions and recommends the Siri exclusion for macOS 27 users. The Mac setting is especially important because installing Signal does not, by itself, provide the same app-enforced defense described for mobile. (Signal)
I’d leave the mobile protections enabled unless there is a specific reason to take a screenshot. The tradeoff is concrete: screenshots are convenient, but disabling the restriction can expose messages from everyone participating in the conversation, including contacts who have made different privacy choices on their own devices.
On Android, I’d also check Screen Security deliberately instead of assuming every phone behaves identically. Signal describes a defense constrained by what it can detect, and Android vendors do not give the app uniform visibility into capture attempts.
Signal’s iOS 27 and Android changes address what happens after a message reaches a participant’s device and becomes readable on screen. End-to-end encryption protects the conversation between participants; a screen-reading agent operates at an endpoint where the message is already being displayed.
That distinction explains why an encrypted messenger would need screenshot restrictions at all. An assistant does not have to break Signal’s encryption to capture the text a person can see. If the operating system supplies that screen content to an agent, the resulting copy can leave the app’s protected environment.
Signal says captured information can become available to the local database or remote server an agent uses. The concern therefore includes local processing, storage and subsequent access, not only whether a screenshot immediately gets uploaded to a cloud service. (Signal)
Apple’s Onscreen Awareness with Siri AI and Google’s Gemini Intelligence are examples Signal names. For readers following Google Gemini’s development, the consequential change is the assistant’s position inside the operating system: it can gain context from other applications, including applications built around strong confidentiality promises.
Useful screen context and message privacy can conflict here. An assistant may need to understand what is visible to answer a question or perform a task. Signal’s position is that a privacy-sensitive app needs a dependable way to refuse that access.
My view is that the important boundary is the app’s ability to control disclosure. A person choosing to send a specific message to an assistant is making an identifiable sharing decision. Background or broadly permissioned access makes that decision harder to see, particularly for the other people whose messages appear in the conversation.
Where Signal’s AI Protection Stops
Signal’s screenshot restrictions on iOS 27 and supported Android devices cover one capture path, not every way an agent could obtain application data. The company specifically identifies filesystem access as another potential route: content already present on a device could be retrieved without taking a screenshot. (Signal)
Three limits shape how readers should interpret the update:
Android protection depends on detection. Signal cannot reliably identify every screenshot attempt across vendors and configurations.
macOS requires user action. Signal recommends excluding the app from Siri because it cannot implement equivalent protection itself.
Blocking the display does not control every data interface. Files, notifications, APIs and other operating-system frameworks need their own access controls.
The practical implication is that a blocked screenshot is evidence of a blocked capture route, not a guarantee against every form of agent access. That is analysis of the documented limits, not a test result.
The update also supplies no evidence that Apple, Google, Microsoft or Meta improperly retained a particular user’s Signal messages. Signal is describing a technical risk and its defensive response.
The Verge’s coverage likewise emphasizes that the remedy is incomplete and that restoring screenshots may expose conversations to AI agents. The reporting documents Signal’s warning; it is not an independent measurement of how often each assistant captures messages.
What Signal Wants From Apple, Google and Microsoft
Signal’s November 20 announcement asks Apple, Google and Microsoft for developer controls that reach beyond screenshot blocking. The company says developers currently face a choice between protecting sensitive data and compromising ordinary app features, such as a user’s ability to take a screenshot. (Signal)
Signal requests two changes:
A documented app-level opt-out from AI data access and processing. Signal wants this to cover screen reading, filesystem access, notifications, APIs and other operating-system frameworks without undermining the app’s core functions.
Accurate technical documentation. Developers need specifications explaining how agents obtain and process application data.
A user-facing Siri exclusion and a developer-enforced opt-out solve different problems. The former depends on each person finding and maintaining a setting. The latter would let a privacy-focused application consistently declare which agent access it permits. That distinction is why the macOS workaround is useful but falls short of Signal’s broader request.
Signal also names Meta’s Muse when discussing third-party agents. It warns that agents can use broad accessibility permissions, originally intended to help people operate their devices, to reach application content. The wider AI-agent ecosystem therefore matters alongside the assistants supplied by operating-system vendors.
The announcement prompted a substantial Hacker News discussion. Commenter nkrisc noted that recipients have always been able to leak conversations. Forgeties79 argued that opaque agent permissions introduce a different problem: participants may not understand what is being shared. Those are individual reactions, not representative testing, but they identify the distinction behind Signal’s response: reducing automated exposure is still worthwhile even when an app cannot prevent every recipient from disclosing a message.
Frequently Asked Questions
3 questions
1
Can I Still Take Screenshots in Signal?
Yes. Signal says users can disable its screenshot protection in the app’s settings. Doing so restores screenshot capability but may also allow AI agents to capture displayed conversations. Signal warns that the consequences extend to your contacts’ messages, so leaving protection enabled is the safer privacy choice. (Signal)
2
Does Signal Block Every Way AI Can Access Messages?
No. Signal explicitly describes its remedy as incomplete. Screenshot blocking does not address every possible route, including an agent retrieving content through filesystem access. Android protection also depends on detecting the relevant access attempt, so the announcement should not be read as a universal guarantee against AI access. (Signal)
3
How Do I Exclude Signal From Siri on macOS 27?
Add Signal to Excluded Apps under Siri in macOS 27’s System Settings. Signal recommends this manual step because it cannot reliably implement the equivalent protection itself on macOS. The exclusion addresses Siri access; it is not a claim that every other potential agent-access route is blocked. (Signal)