Signal President Meredith Whittaker calls out agentic AI as having ‘profound’ security and privacy issues
Signal President Meredith Whittaker warned on Friday that Agentic AI could bring a risk for users’ privacy.
On the stage on the SXSW conference in Austin, Texas, the advocate for safe communication to the use of AI agents and “put your brain in a pot” and warned that this new paradigm of computing – where AI performs tasks – has a “deep problem” on behalf of users with both privacy and security.
Whittaker explained how AI agents are put on the market as a way to add value to your life by processing different online tasks for the user. AI agents, for example, would be able to take on tasks such as looking up concerts, book tickets, plan the event in your agenda and sent your friends messages that it has been booked.
“So we can just put our brains in a pot because the thing does and we don’t have to touch it, right?”
Then she explained access to the AI agent that should perform these tasks, including access to our web browser and a way to stimulate them, as well as access to our credit card information to pay for tickets, our agenda and messages to send the text to your friends.
“It should be able to drive that [process] Over our entire system with something that looks like root permission, access to each of those databases – probably in it clear, because there is no model to do that, “warned Whittaker.
“And if we are talking about a sufficiently powerful … AI model that drives that, does not happen on device in any way,” she continued. “That is almost certainly sent to a cloud server where it is processed and sent back. So there is an in-depth problem with safety and privacy that chases this hype around agents, and which ultimately threatens to break the blood-brain barrier between the application layer and the OS layer by processing all these individual services [and] Muddyden their data, “concluded Whittaker.
If a messages app such as signal were integrated with AI agents, this would undermine the privacy of your messages, she said. The agent must gain access to the app to text your friends and also withdraw data to summarize those texts.
Her comments followed comments that she previously made during the panel about how the AI industry was built on a security model with mass data collection. She said that “larger is better AI paradigm” – which means that the more data, the better – had possible consequences that she didn’t think was good.
With Agentic AI, Whittaker warned that we would further undermine privacy and security in the name of a “Magic Genie Bot that will take care of the existence of life,” she concluded.
