Signal has taken a significant step to enhance user privacy by default, blocking screenshots to prevent AI agents from accessing private conversations on Android and iOS devices. Despite the move, Signal acknowledges that this measure provides only partial protection.
New Privacy Measures on iOS and Android
This new feature targets AI systems integrated into iOS 27 and compatible Android versions, aiming to detect and block attempts by AI assistants to read Signal screens. This update builds on a similar function introduced for Windows 11, following Microsoft’s controversial Recall feature.
Meredith Whittaker, the president of the Signal Foundation, described the integration of AI into various operating systems as an “existential threat” to Signal’s privacy commitments. By scanning the information on screens, these AI agents can potentially expose Signal data, undermining the platform’s privacy guarantees.
Risks of AI Access to Private Data
The concern is that AI agents could capture and store Signal information on local or remote servers outside the platform’s control, posing a significant privacy risk to users. While users can disable this feature through their device settings, doing so may expose their Signal data to AI agents, compromising privacy.
Signal admits that restricting screenshots does not fully prevent AI agents from accessing application data through other means, such as file system access. The company also highlights the lack of effective tools for developers to prevent such privacy intrusions by AI agents.
Call for Action from Tech Companies
Signal urges tech companies to provide two straightforward solutions: allow apps to opt out of AI data access, including screen reading, and improve documentation on AI agent functioning concerning data access. These measures aim to empower developers to design effective privacy controls.
Recent reports by Reuters underscore the urgency, revealing instances where Meta’s AI agent, Muse, accessed Signal messages on various devices by exploiting system permissions. Such incidents highlight the need for robust tools to prevent unauthorized access by AI agents.
Apple’s recent iOS 27.2 update and Google’s Gemini settings offer users more control over app access, reflecting a growing trend towards enhancing privacy in the AI era. These updates provide a glimpse into the future of privacy settings aimed at protecting user data from AI agents.
