Signal says operating-system AI assistants threaten app-level privacy because they can inspect what appears on a screen, then store or process sensitive content beyond an app’s control. Even though Signal’s open-source software and verified encryption protect messages in transit, they cannot stop an operating system from capturing messages after they are displayed. Signal compares this risk to Microsoft Recall, which records screenshots and creates a searchable database that could retain even disappearing messages.
To limit exposure, Signal will block screenshots in iOS 27 and, on Android devices where it can detect an assistant reading the screen, enable its Screen security setting. Users can turn off the protection, but doing so may expose their own and their contacts’ messages; the safeguards also cannot prevent every route of access, such as filesystem retrieval. Signal says it lacks reliable tools to apply comparable protections on macOS and many Android variants, and that poor documentation has forced developers to investigate how these systems work themselves. It urges Apple, Google, and Microsoft to provide clear technical specifications and straightforward controls letting apps opt out of AI access across screens, files, notifications, and other system features without sacrificing usability. Signal also warns that third-party agents may exploit accessibility permissions intended for people, and calls for industry-wide protections rather than app-specific workarounds.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.