Signal to block screenshots again as AI agents raise privacy fears
Signal is stopping screenshots on its app due to worries that new AI agents in phones could compromise user privacy. The company says this is needed to protect sensitive chats.
Signal is turning off screenshots in its app once more, this time because of concerns that AI agents built into phones are threatening user privacy.
The encrypted messaging service will soon stop screenshots on iPhones running version 27 of its app, and on Android devices when AI assistants try to grab screen data. Signal President Meredith Whittaker announced the move in a blog post, explaining that the company is responding to a wave of new AI-powered features from Apple and Google, such as Siri AI’s Onscreen Awareness and Google Gemini Intelligence.
Privacy at risk from AI
These AI features are designed to watch what’s on a phone’s screen so they can answer questions or complete requests. But Whittaker said they can also scoop up anything visible in an app, including private Signal messages.
“Your sensitive data is being captured, stored, and made accessible to whatever local database or remote server an OS-level AI agent is using at that moment,” said Whittaker. “All of this is outside of Signal’s control once the data has been captured, and imperils the privacy of all the participants in a conversation.”
The changes mean users won’t be able to take screenshots by default. If someone wants to take a screenshot anyway, they can turn the setting off, but Signal warned this puts both the user and their contacts at risk.
Signal first blocked screenshots last year after Microsoft released its Recall AI feature. That feature could take snapshots of everything on a Windows 11 device, including messages inside Signal. Now, with Apple and Google rolling out their own agents and Samsung reportedly working on a competing tool called Galaxy Claw, privacy-focused apps like Signal say they have to adapt fast.
Other risks and limits
Whittaker admits that blocking screenshots isn’t a full solution. She said AI agents might still get access to app data in other ways, such as by reading files from the device storage. Earlier in the week, a Reuters report, supported by Signal and the AI Now Institute, found that Meta’s Muse AI agent used setup permissions to get into apps and data users hadn’t clearly given it access to.
Signal’s new protections do not cover Mac computers. Whittaker said Signal and other app makers don’t get strong tools from Apple, Google, or Microsoft to stop AI agents from poking into private data, and that blocking AI access often means making big sacrifices to the user experience.
“It is a serious problem that Signal, and application developers beyond, lack robust tools to prevent AI agents from undermining application-level privacy,” Whittaker said.
Last week, Google DeepMind researchers used AI to formally check that Signal’s encryption still works as designed, something the company sees as a high bar for security. But Whittaker pointed out that verification doesn’t help if an AI agent can simply copy messages from the screen or device itself.
Signal is advising users to keep screenshotting disabled to maintain the privacy of their conversations. The company says it will keep looking for ways to protect chats as the use of AI agents grows.
Tagged Culture