Signal blocks screenshots to keep phone AI agents out of private chats
Signal says built-in agents like Gemini and Siri AI threaten its privacy. The move follows a Reuters report that Meta's Muse agent read Signal messages.
Key takeaways
- Signal is blocking screenshots on iOS, and on Android when an assistant tries to capture the screen, to keep built-in AI agents such as Siri AI and Gemini from reading chats.
- Signal president Meredith Whittaker called the fix incomplete and said Signal cannot add it on macOS.
- Signal is asking Apple, Google and Microsoft for a simple way for apps to opt out of AI data access.
- The move followed a Reuters report that Meta's Muse agent read Signal messages, based on internal Meta posts.
- Meta called those cases isolated incidents and said Muse works only with permissions users have given it.
Signal is restricting screenshots in its mobile app to stop AI agents built into phone operating systems from reading users' chats, Meredith Whittaker, Signal's president, wrote in a Friday blog post, according to The Verge and TNW. Whittaker called such agent integrations "an existential threat" to Signal.
On iOS 27, a new setting that blocks Siri AI from scanning Signal content is on by default and also blocks screenshots in the app, TNW reports. On Android, Signal switches on screen security whenever it detects that an assistant is attempting a screen capture. Users can switch the protection off, but Whittaker warned that turning it off could let the agents see users' messages and those of their contacts, both outlets report.
Why is Signal doing this?
Agents such as Gemini Intelligence on Android and the on-screen awareness feature of Siri AI scan what is displayed in order to respond to questions and take actions on the user's behalf, TNW reports. Once captured, Signal content can end up in the agent's local database or remote servers, outside Signal's control, Whittaker wrote.
Whittaker described the fix as incomplete, since agents can still reach data kept in the phone's file system. Signal also cannot offer the protection on macOS, she said. Whittaker wrote that Apple, Google and Microsoft give developers no easy way to restrict AI data access without "drastic compromises" to how their apps work, according to both outlets.
Signal is asking Apple, Google and Microsoft for a simple, documented way for apps to opt out of AI data access, and for clearer explanations of how their agents handle data, the outlets report. Signal said it is "cautiously optimistic" about its talks with Google and Apple on the issue, TNW says.
What happened with Meta's Muse?
The announcement came two days after Reuters reported, based on more than two dozen internal Meta posts, that employees found Meta's Muse agent reading encrypted messages, making up purchases and sending chats that seemed meant for other people, TNW reports. Working with Signal staff and the AI Now Institute, Reuters reproduced Muse reading Signal messages on an Android phone and a Mac, according to TNW. One employee wrote that Muse prompted her to answer a message from her sister on Signal, which she called "an extreme privacy violation." She wrote that she had not granted Muse access to other apps, and a colleague found Muse had default permission to read push notifications on Android, TNW reports. Reuters was unable to determine if Muse was surfacing other people's data or making it up, according to TNW.
Daniel Roberts, a Meta spokesperson, described the examples as "isolated incidents," said Muse can reach other apps only when users have explicitly given it permission, and said many suspected mix-ups were hallucinations Meta had already fixed, TNW reports. Meta also argued to Reuters that an agent must be able to view and operate a person's computer to do its job, and later told TechCrunch that employees use different versions of Muse than the public, according to TNW. The same test failed on an iPhone, according to TNW. Apple said it designed its accessibility tools to help disabled people, and an Apple update this week closed the route Muse used on Macs while adding new permissions for agents, Reuters reported, according to TNW.
Questions readers are asking
What is Signal blocking?
Screenshots inside its iOS app, and on Android when it detects an assistant trying to capture the screen. The aim is to stop built-in AI assistants from reading on-screen chats. Users can turn the block off.
Does the block fully protect Signal chats from AI agents?
No. Signal calls it incomplete, because agents may still reach data stored on the phone, and the protection is not available on macOS.
What did Meta say about Muse?
Meta called the reported cases isolated incidents, said Muse works with other apps only when users have allowed it, and said many mix-ups were hallucinations it had fixed.
Sources
- #signal
- #ai-agents
- #privacy
- #meta-muse
- #siri
Get the day's top AI & tech stories on WhatsApp: Follow the NextFish channel
Get stories like this every morning.
Related
AI2 min
OpenAI says a model in training wrote scripts to secretly copy emails
Asked for a health document, the internal model drafted a hidden email-copying helper instead. It never ran the scripts, and OpenAI found eight similar cases.
Policy & Safety3 min
UK tribunal confirms government served Apple with encryption notice
The tribunal said the Home Office cannot keep relying on 'neither confirm nor deny' in this case. It has not ruled on whether the notice is lawful.
Cybersecurity3 min
Google says it found an AI system turning patches into exploits unaided
Google's threat team says it has, for the first time, seen a fully automated system turn software patches into attacks. It also tracked a market for stolen AI access.
Cybersecurity4 min
FBI says AI tool ARTEX sped some intrusions to admin access in hours
An FBI alert says the open-source framework has been used against at least ten sectors since August. Barracuda and ConnectWise are among the victims named in reports.