New Today USA

Signal Blocks Screenshots to Stop AI Agents Reading Your Private Messages

Signal, the encrypted messaging app, has announced that it will block screen captures by default on Android and iOS, in a move specifically designed to stop AI agents built into phones from reading users’ private messages.

The change, announced on 20 November, is a direct response to a new kind of privacy risk. Phones now come with AI assistants built in, such as Gemini on Android and Siri on Apple devices, and those assistants are increasingly able to see what is on your screen. That can be useful when you want help with something you are looking at, but it also means that an AI system could, in principle, read your private messages, your photos, and other sensitive information on your screen, even if you never asked it to.

Signal’s answer is simple: stop the screen from being captured in the first place. By blocking screenshots and screen recordings by default, the app makes it much harder for an AI agent, or any other software on the phone, to take a picture of your messages and read them. It is a technical fix to a problem that did not exist a few years ago, and it shows how quickly the privacy landscape is changing.

For Signal, this is not a surprising move. The app has built its reputation on privacy. Unlike many other messaging apps, Signal uses end-to-end encryption by default, which means that only the sender and the recipient can read a message. Not Signal itself, not a government, not a hacker who intercepts the message in transit. That encryption has made Signal popular with journalists, activists, and anyone who wants to keep their conversations private.

But encryption only protects a message while it is travelling between phones. Once it arrives on your screen, it is visible, and if something on your phone can see your screen, it can see your message. That is the gap Signal is now trying to close. The company has long offered a feature to block screenshots, but making it the default setting is a much stronger step. It means that most users will be protected without having to change anything, and without even knowing the setting exists.

The timing of the announcement matters. AI agents are becoming more powerful and more deeply integrated into phones. They are no longer just voice assistants that answer questions. They are systems that can look at your screen, understand what you are doing, and offer to help. That can be genuinely useful. An AI agent could summarise a long email, help you fill in a form, or remind you about something you were reading. But it also raises a question: how much should an AI agent be allowed to see?

For most people, the answer depends on what is on the screen. If you are looking at a recipe, or a news article, or a public web page, there is not much harm in an AI agent seeing it. If you are looking at a private message from a friend, a photo from your family, or a conversation about something sensitive, the calculation is very different. You did not invite the AI agent into that conversation, and you may not want it there.

Signal’s move is a recognition of that difference. It is saying that private messages should stay private, even from the AI systems built into your own phone. It is drawing a line between the phone as a tool you control, and the phone as a device that watches what you do and reports back to an AI system.

That line is becoming harder to draw. Phone makers are competing to build the most helpful AI assistants, and the most helpful assistant is often the one that knows the most about you. It can only help you if it can see what you are doing, read what you are reading, and understand what you need. That creates a tension between usefulness and privacy, and different companies are resolving it in different ways.

Some companies argue that AI agents only read your screen when you ask them to, and that they do not store what they see. Others argue that the data stays on your phone and is never sent to the cloud. Those reassurances may be true, but they are hard for an ordinary user to verify. You cannot see what an AI agent is doing in the background, and you cannot be sure that a setting you changed months ago is still in place.

Signal’s approach is different. Instead of asking users to trust that an AI agent will behave itself, it removes the opportunity altogether. If the screen cannot be captured, the AI agent cannot read it. It is a blunt solution, but it is also a clear one, and it fits with Signal’s philosophy of building privacy into the technology itself, rather than relying on promises and settings.

Those questions are not going away. AI agents are only going to become more capable, and more deeply embedded in our phones, our computers, and our homes. The more they can see, the more they can help, but the more they can see, the more they can learn about us, often without us realising it. That is a trade-off that every user will have to think about, and every company will have to address.

Signal’s announcement is a reminder that privacy is not just about encryption, or about keeping hackers out. It is also about keeping the systems on your own phone from watching you. It is about who gets to see your messages, and when, and whether you get a say in it. For Signal, the answer is clear: your private messages are yours, and no AI agent should be reading them without your permission.

Whether other messaging apps follow Signal’s lead remains to be seen. Some may argue that blocking screenshots is too extreme, or that it will annoy users. Others may see it as a selling point, a way to show that they take privacy seriously in an age when AI is watching more and more of what we do. Either way, Signal has started a conversation that needed to happen, and it has done so in the most practical way possible: by changing the default, and by making privacy the easy choice.

Exit mobile version