Signal Introduces Screenshot Restrictions to Prevent AI Agents from Reading Private Messages
Signal is rolling out a new setting enabled by default that prevents AI agents from scanning private messages by blocking access to phone screenshots. When screenshot activity or an AI assistant attempting to access the screen is detected, users will receive a pop-up and can choose to lift the restriction. Signal is also calling on operating system providers to offer clearer opt-out mechanisms and disclose what data AI agents can access. The company says screen reading could transfer end-to-end encrypted messages to other servers, thereby bypassing the app's own privacy protections.
Why read this
SelectedAI score 76
Signal's new setting offers a concrete approach for privacy apps to prevent system-level AI agents from reading on-screen content, while also revealing the limits of existing permission mechanisms.
Topics
Body · AI translation
Signal has long been regarded by privacy advocates as a leader in secure communications. The company said the new protections being introduced are incomplete and called on operating system providers such as Apple, Google parent Alphabet, and Microsoft to give app developers more specific and effective ways to prevent AI agents from accessing sensitive data.
Signal also asked these companies to disclose more about how their AI agents operate, enabling developers to assess what data the programs are accessing and adjust their own systems accordingly. The move reflects growing tensions between privacy-focused apps and major technology companies, which are racing to launch AI agents capable of viewing users’ screens and taking actions on their behalf based on what they see.
Signal specifically said that AI capabilities built directly into operating systems, such as Apple’s Siri AI and Google’s Gemini Intelligence, are particularly concerning because of their deep integration with users’ devices. The company said operating systems provide the environment in which Signal and other apps run, and the way this “agentic AI” is being integrated is making that environment increasingly hostile to privacy. Signal called this an “existential threat” to the company and its ability to responsibly fulfill its mission in a blog post provided to Reuters.
To prevent AI agents from scanning Signal messages, the service is rolling out new settings that block access to screenshots on phones, with the restrictions enabled by default. When Signal detects an attempt to take a screenshot or an AI assistant trying to access the screen, users will receive a pop-up notification and, depending on their device, can choose whether to lift the restrictions.
Signal blocked Microsoft’s Recall AI from taking screenshots on Windows last year. The company said it cannot reliably provide similar protection on Apple’s macOS.
AI agents capable of taking actions on users’ behalf have been promoted as the next major wave of AI technology. They can pull relevant information from different apps and independently complete time-saving tasks such as replying to emails, scheduling appointments, and shopping online, without requiring users to provide step-by-step instructions. However, this capability may also bypass protections that apps such as Signal have built for themselves.
Reuters reported earlier this week that Meta’s Muse AI agent could bypass its official “Connector” tool to access Signal messages on Apple computers and phones running Google’s Android operating system. In both cases, Muse obtained the messages by using broad operating system-level permissions granted during the app’s initial setup, employing those permissions in ways users might not have anticipated.
On Apple’s macOS, for example, Muse combined the computer’s accessibility features with screenshot tools, repurposing settings Apple provides for people with disabilities to navigate to the Signal interface, take screenshots of chats, extract text from the images, and reproduce the messages within the Muse app. A system update released by Apple this week effectively closed this specific route, but the company also introduced a new set of permissions for AI agents that provide similar capabilities, accompanied by additional user warnings.
Signal said developers should have a direct way to allow apps to opt out entirely from data processing by AI assistants, particularly regarding screen reading, notifications, and file system access. As a service owned by a nonprofit foundation, Signal promises users a secure communications experience by limiting the data stored on its servers, refusing advertising, and providing end-to-end encryption, meaning the company cannot access or read users’ conversations within its app.
Signal said AI systems capable of scanning users’ screens could extract messages from Signal’s secure environment and transmit them to other servers used by the agent, thereby undermining those protections.
AI Trends