Signal restricts OS-level AI agents from capturing its data on iOS 27 and Android

Signal Technology Foundation, the nonprofit behind the Signal app, announced in a blog post that it is restricting the latest AI agents built into mobile operating systems from potentially capturing and storing Signal data. The post is signed by the organisation; no individual author is named.
Signal has done this once before. Last year, it used screenshot blocking to stop Microsoft's Recall AI from taking and storing screenshots of Signal messages on Windows 11 by default. A footnote explains why: Recall captures the screen every few seconds, extracts text from the screenshots with optical character recognition, and fills a searchable on-device database that the AI system can query, potentially exposing and retaining Signal data indefinitely, disappearing messages included.
Now Signal is taking similar measures against agentic systems built into iOS 27 and, on Android, on versions where it can detect an AI assistant trying to read the Signal screen. The trigger is that Google and Apple are integrating agentic features, named in the post as Gemini Intelligence with Google Gemini and Onscreen Awareness with Siri AI, deep into their operating systems. Signal argues that Google (Android), Apple (iOS/macOS) and Microsoft (Windows) make the three major consumer operating systems, the environment in which third-party apps must work, and that these integrations make that environment increasingly hostile to privacy. It calls this an existential threat to Signal and its ability to uphold its mission.
The mechanism, as Signal describes it: to learn about the user, OS AI agents "scan" what is on the screen. That captures Signal data and hands it to the agentic systems so they can answer questions and act for the user. Signal says this undermines its ability to guarantee privacy independently, a guarantee backed for over a decade by its open source code base and formally verified encryption protocol. Once data is captured, it may sit in whatever local database or remote server the OS-level agent uses, outside Signal's control, which imperils every participant in a conversation.
The fix is to block capture: Signal prevents screenshots in the app on iOS 27, and on Android blocks capture when AI assistants attempt to read screen data. A footnote adds detail for Android. There are many Android versions beyond Google's own distributions on Pixel devices, and for many vendors Signal cannot reliably detect a screenshot attempt when screenshots are already disabled. Rather than make screenshots simply stop working and confuse users, Signal chose to detect AI assistant access to the screen, and it automatically enables the "Screen security" setting in Signal Android when that access is detected.
Users who want to take screenshots can turn the protection off in Signal's settings. Signal warns that doing so will potentially expose their and their contacts' Signal content to these AI agents and break its privacy guarantees. It also calls its own remedy incomplete: restricting screenshots does not stop other, sneakier routes, such as an agent retrieving content already on the device through the filesystem.
macOS is not covered. Signal says Google, Apple, Microsoft and others do not give developers easy ways to restrict AI data access without drastic compromises to core user experience, which is also why it cannot implement the protection on macOS. A footnote says Signal cannot reliably block screenshots there. Protections against continuous screen capture exist on macOS but cannot be applied to screenshots, which is how most agentic systems get data from apps like Signal. Signal describes a minefield of bewildering decisions, including APIs deprecated in favour of screen-protection solutions that only work for streaming media and other DRM-protected content, and says it hopes Apple fixes this urgently. For now, its only recommendation for macOS 27 users is to add Signal to the Excluded Apps section in System Settings > Siri, to keep message content from proactive access by Siri AI.
Signal also complains of a lack of documentation on how these agents work. Over the past few months its developers have spent considerable time and energy on forensic investigations across Android and iOS to learn what happens when these systems access app data. It says that without better documentation and clear technical specifications, it will be increasingly hard to protect the many millions of people who rely on it.
Signal calls on Apple, Google and Microsoft for two remedies. First, developers should have an easy, well-documented way to opt their app out of AI data access and processing, including screen reading, filesystem access, notifications, APIs and other OS frameworks, without undermining the app's core utility. Second, OS companies should publish better documentation and technical specifications on how their AI agents work, particularly on data access and processing. Signal adds that confidential computing approaches are not a complete solution, because they still leave the power to decide the extent of privacy guarantees (if any) with the providers.
The post also sees a growing need for tools against third-party agents, such as Meta's Muse, that use the deep permissions users are encouraged to grant them during setup. In many cases, Signal says, these agents lean on accessibility settings that were meant to help people use their apps, not to act as a skeleton key for AI agents.
Signal says it is talking with the wider ecosystem and is cautiously optimistic that some major OS providers, like Google and Apple, are engaging with it on how these systems should interface with privacy-sensitive apps. It hopes to have more to share soon and thanks allies pushing for similar changes. It closes by calling itself the canary in the coal mine.
Key facts
- Signal blocks screenshots in its app on iOS 27, and on Android it automatically turns on its "Screen security" setting when it detects an AI assistant trying to read the screen.
- The move follows last year's use of screenshot blocking against Microsoft Recall on Windows 11, which Signal says it did by default.
- Signal calls its fix incomplete: AI agents could still potentially reach app data through the filesystem, and macOS is not covered, where Signal only advises adding it to Excluded Apps in System Settings > Siri.
- Signal asks Apple, Google and Microsoft for an easy, documented way for developers to opt an app out of AI data access, and for better documentation of how the AI agents work.
- Users can switch the protection off in Signal's settings, but Signal warns this potentially exposes their and their contacts' content to AI agents.
Why it matters
Signal's argument is that its privacy guarantee, backed for over a decade by open source code and a formally verified encryption protocol, can be undone from outside the app. If an OS-level AI agent scans the screen, message content is captured and passed to whatever local database or remote server that agent uses, beyond Signal's control. Signal calls this an existential threat and says it affects the whole app ecosystem, not only Signal. The post is also a public demand: the three big OS makers should give developers an opt-out and real documentation.
Who it affects
Signal users on iOS 27 and on Android versions where Signal can detect an AI assistant reading the screen, and the contacts in their conversations, since one person's setting changes what happens to everyone's messages. macOS 27 users get only a manual workaround. Other app developers are affected too: Signal says they lack robust tools to stop AI agents from undermining app-level privacy. Apple, Google and Microsoft are the named addressees of the demands, and Meta's Muse is named as an example of a third-party agent.
How to use it
On iOS 27 the screenshot blocking is applied in the app. On Android, Signal turns on its "Screen security" setting automatically when it detects AI assistant access. Anyone who wants to take screenshots can disable the protection in Signal's settings, with the exposure Signal describes. On macOS 27, add Signal to the Excluded Apps section in System Settings > Siri to keep message content from proactive access by Siri AI. Developers of other apps are not offered a tool here; that is the gap the post asks the OS makers to close.
How solid is it
This is a first-party statement from Signal Technology Foundation, so the account of what Signal is doing is direct. The post speaks of potential capture and gives no evidence that a specific OS AI agent has actually leaked or misused Signal data. It does not give release dates or shipping status for iOS 27, macOS 27 or the named AI features, and no response from Apple, Google, Microsoft or Meta is quoted. Signal says only that some providers, like Google and Apple, are engaging with it. It does not list which Android versions or vendors the detection works on beyond "versions where we can detect an AI assistant attempting to read your Signal screen".
Risks and caveats
Signal itself calls the remedy incomplete: blocking screenshots does not stop agents that retrieve content already on the device via the filesystem. On Android, detection is imperfect because many vendors give no reliable way to detect a screenshot attempt when screenshots are already off, so Signal watches for AI assistant access instead. macOS is not protected, because Signal says it cannot reliably block screenshots there. Turning the protection off exposes content and breaks Signal's privacy guarantees. Signal also says confidential computing approaches are not a complete solution. No timescale is given for further news or for macOS protection.
“Signal is the canary in the coal mine, but this affects the entire ecosystem and should galvanize anyone who cares about these fundamental rights.”
— Signal Technology Foundation, Signal blog post