Signal is now blocking the AI assistants built into iOS 27 and Android from reading its chats, because an AI agent built into a phone's operating system can see whatever your team sees on screen. If client contracts, invoices or personal data pass through messengers on work phones, decide now which devices and apps should run with OS assistants switched off.
According to Signal, the app will block screenshots on iOS 27 and stop capture on Android wherever it can detect an assistant trying to read the screen. Last year it did the same against Microsoft Recall on Windows. The reason is that features like Gemini on Android and Siri's Onscreen Awareness scan the screen. That data then lands in whatever local database or remote server the assistant uses, outside the app's control.
What this means for business automation
The protection has gaps, and Signal says so itself. macOS is not covered, and screenshot blocking does nothing against agents reading files already stored on the device. So the practical rule for automation is about architecture. An agent that handles client data should receive it through an API with a narrow scope, such as a CRM endpoint, a mailbox or a document folder, rather than by watching a screen. In the Claude-based pipelines I build, the model gets only the fields a task needs, and logs stay on the client's server. This is one of the data security questions clients ask before a project starts.
Why it matters: Built-in AI assistants are becoming a new path for client data to leave the business, and app developers cannot fully close it yet. Keep sensitive conversations off devices with screen-reading assistants enabled, and give your own agents data only through explicit, limited integrations.
Source: Hacker News