What changes and why

Signal has announced that it has introduced restrictions on screenshots to prevent AI agents from reading private chats. According to the source, the app blocks screen captures on iOS 27 and activates protection on supported Android versions when an AI assistant tries to access the screen. The change is not a new encryption feature but an attempt to close a practical gap: decrypted messages displayed on the screen can be captured and taken outside the app by tools that operate at the operating-system level.

The mechanism matters because assistants integrated into systems, such as Apple's Siri AI and Google's Gemini Intelligence, can capture the content of the screen to complete tasks requested by the user. In this way, according to Signal, end-to-end encryption is bypassed not through a cryptographic attack but by reading what is visible on the screen. Signal described the situation as an existential threat to privacy, referring to the consequences of this mode of access.

The Meta case and the limits of the defenses

The source cites a specific episode: Meta's AI agent, called Muse, allegedly captured Signal content on Android and macOS by exploiting broad configuration permissions, including accessibility tools and screen capture on macOS. Meta stated that this occurred only with the user's explicit consent. This is an important clarification, because it shifts the discussion to the ground of permissions granted by the user and how operating systems manage them, not necessarily to unauthorized access.

Signal acknowledges, however, that its defenses remain incomplete. Screen captures on macOS cannot be reliably blocked, and AI agents' access to the file system and notifications remains a risk. The app had already blocked Microsoft Recall on Windows, but the new measure does not cover all the channels through which an assistant could read private information. This admission is central: the announced protection reduces an exposure surface, but it does not eliminate the possibility that decrypted content or metadata ends up outside the app.

The request to operating system makers

Signal is asking Apple, Google, and Microsoft to make it easy for apps to block AI agents' access entirely. The request shifts the burden from the individual application to the platform makers, who control the rules for screen access, accessibility permissions, and notifications. If protections remain entrusted to isolated app initiatives, the level of protection may vary depending on the operating system, version, and settings chosen by the user.

Several aspects that the material does not clarify remain to be verified: which Android versions are affected by the protection, how exactly the block behaves on iOS 27, and whether the announced measures are already active or have only been announced. It is also not specified whether Apple, Google, or Microsoft have responded to Signal's request. The news, attributed to The Decoder, should therefore be read as a description of an announced measure and of a critical issue declared by the app, not as the definitive solution to the problem of AI agents' access to private communications.

Source and attribution: this original summary was prepared by AICronaca based on the report by The Decoder ↗. The statements are attributed to the source; any information limitations: The material does not specify which Android versions are involved, the actual release status of the measures, or any responses from Apple, Google, and Microsoft. The Meta case is reported with the company's clarification on user consent, but without technical details that can be independently verified..