Signal President Meredith Whitaker warned on Friday that the Agentic AI could take the chance of the user's privacy.
At the SXSW conference in Austin, Texas, on stage, the lawyer for secure communication described using AI agents as “their brain in a glass”, and warned that this latest paradigm of the pc – by which AI performs tasks in use – with each privacy and security.
Whittaker explained how AI agents are marketed with a purpose to create their lives through the transfer of various online tasks for the user. For example, AI agents could have the ability to tackle tasks equivalent to doing concert events, booking tickets, planning the event of their calendar and sending their friends who’re booked.
“So we will simply put our brain in a glass since the thing does it and we don't must touch it, right?” Throughout Whittaker.
Then she explained the kind of access that the AI agent needed to perform these tasks, including access to our web browser and the likelihood to advance you, in addition to access to our bank card information to pay tickets, our calendar and messaging app to send the text to your pals.
“It should have the ability to advance this (process) in our entire system with something that appears like the basis permit and accesses each individual of those databases – probably because there isn’t any model to make this encrypted,” warned Whittaker.
“And after we discuss a sufficiently powerful … AI model that’s writing this, there isn’t any option to pass on the device,” she continued. “This is sort of definitely sent to a cloud server on which it’s processed and returned. So there may be a deep problem with certainty and privacy that pursues this hype about agents and ultimately threatens to interrupt the blood-brain barrier between the appliance layer and the OS layer by disguising all of those separate services (and), ”concluded Whittaker.
If a messaging app like Signal integrated into AI agents, it could undermine the privacy of her news, she said. The agent must access the app to write down your pals and withdraw data to summarize these texts.
Her comments followed the comments that she previously made in the course of the committee about how the AI industry was built on a surveillance model with mass data acquisition. She said that the “greater a greater AI paradigm” -ie more data, the higher -had the higher consequences that she didn't think was good.
With the Agentic Ai, Whittaker warned that within the name of a “magical geniusbots that takes care of the necessities of life”, we’d also undermine privacy and security.
