Signal blocks screenshots to keep Siri AI and Gemini agents out of users' messages
Signal is blocking screenshots in its app on iOS 27, and on Android whenever it detects an AI assistant trying to capture the screen, to stop AI agents built into mobile operating systems from reading users' messages, the company said on 20 November.
Signal says Google and Apple are building agentic features such as Gemini Intelligence and Siri AI's Onscreen Awareness deep into Android and iOS. These agents "scan" what is on the screen, so Signal data is captured and made available to them, which the company says undermines its ability to guarantee privacy. It took a similar step last year against Microsoft's Recall on Windows 11.
Users who want screenshots can turn the protection off in settings, at the cost of exposing their and their contacts' messages to these agents. Signal calls its remedy incomplete: it cannot cover every Android device, cannot reliably block screenshots on macOS, and does not stop other ways agents could reach app data. It also points to third-party agents such as Meta's Muse using accessibility permissions as "a skeleton key."
Signal is asking Apple, Google and Microsoft for two things: an easy, documented way for apps to opt out of AI data access, and better documentation of how their AI agents access and process data. It calls the current integration of AI agents into operating systems "an existential threat to privacy at the application layer."
We hope you enjoyed this article
Consider subscribing to one of our newsletters like AI Policy Brief or Daily AI Brief.
Also, consider following us on social media:
AI Policy Brief
Weekly report on AI regulations, safety standards, government policies, and compliance requirements worldwide.
Whitepaper
Tensordyne Napier: What If One Rack Could Do the Work of Nine?
Tensordyne
This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.
Read moreYou may also like
Apple warns AI agents could exceed permissions and conceal actions
Google details privacy protections in its new intelligent glasses
Google DeepMind's AlphaProof Code formally verifies the Signal protocol's cryptographic code in four weeks
Apple could make Siri more like an agent as early as next year, Gurman reports
Chicago Public Schools blocks ChatGPT, Claude and Gemini for students as Google ties Search to AI
Daily AI Brief: the AI news that matters, in your inbox.