Signal is taking measures to isolate its messages from the AI assistants now built into smartphones, restricting the screenshot function to prevent systems such as Google's Gemini and Apple's Siri from quietly scanning and storing users' private conversations.
This non-profit encryption communications organization stated on Thursday that it will block the latest proxy AI feature in iOS. On Android, when the system detects that a AI assistant attempts to read a user's Signal screen, the "screen security" setting will be automatically enabled.
Signal said that this move continues with the actions taken against Microsoft last year by Recall. Recall is a feature of Windows that periodically captures screenshots; Signal prevented it from storing Signal message images by default at that time.
Signal indicates that the issue lies at the operating system level. AI proxies work by "scanning" the content on the user's screen, which means that Signal data will be captured and handed over to these systems for processing.
Once this happens, the content will enter the local database or remote server used by AI, and it will be out of the control of Signal. It also exceeds the scope of encryption protection that this application has emphasized for over a decade.
The president, Meredith Whittaker, stated bluntly that operating system-level AI integration poses a "threat to survival" to this application and believes that Apple, Google, and Microsoft are reshaping the environment in which each application operates in a way that is "increasingly hostile to privacy."
If users still wish to take screenshots, they can disable this protection in the Signal settings. However, the company warns that doing so may expose their messages as well as those of their contacts to AI proxies.
Signal also acknowledges that this fix is not complete: it cannot prevent more concealed methods of access, such as directly extracting data from the device's file system; moreover, it is currently not possible to implement it on macOS either, because Signal states that Apple has not provided a documented tool that can reliably prevent screenshots.
This privacy-conscious company also calls on the three major operating system giants to provide developers with a simple and well-documented way for applications to opt out of AI data access, and to publish clear guidelines on how these proxies actually process data. Signal stated that in order to understand what these systems are actually doing, the company has spent months conducting an “evidence gathering investigation.”
Signal also named Meta's Muse and other third-party proxies, warning that they had abused the auxiliary function permissions that were supposed to help users, turning these permissions into a "universal key" for AI.
This concern is not unfounded: A journalist discovered that Muse accessed his private iMessages when his settings allowed it, and later gave a wrong explanation for how he came to know this information. This incident has raised further concerns about how much data proxies could access once they gain deep device permissions.
The measures taken by Signal are set within a broader context of reflection: the outside world is re-examining how AI proxies access data that they shouldn't, from uncontrolled proxies breaching systems to banks being busy preventing autonomous software from entering sensitive accounts, all of which are part of this trend.












