Signal President Meredith Whitaker calls on AI agency as “deep” security and confidential problems

Signal President Meredith Whitaker warned on Friday that Agent AI may come at the risk of consumer confidentiality.

Speaking on the stage of the SXSW conference in Austin, Texas, the defender of secure communications pointed to the use of AI agents such as “placing your brain in Burkan” and warned that this new paradigm of calculations – where AI performs tasks on behalf of users – there is a “depth problem”.

Whittaker explained how AI agents are offered as a way to add value to your life by processing different online tasks for the user. For example, AI agents will be able to take tasks such as searching for concerts, book tickets, plan the event on their calendar, and send messages to their friends that it is booked.

“So we can just put our brain in a jar, because the thing does it and we don’t have to touch it, right?”

She then explained the type of access that the AI ​​agent will have to perform these tasks, including access to our web browser and a way of managing it, as well as access to information about our credit card to pay for tickets, our calendar and message application to send the text to your friends.

“He should be able to manage this (the process) through our entire system with something that looks like a radical permission, access to each of these databases – probably clear because there is no model to make this encrypted,” Whitaker warned.

“And if we are talking about a powerful enough… AI model that power it, there is no way that happens on the device,” she continued. “This is almost certainly sent to a cloud server where it is processed and sent back. So there is a deep problem with security and confidentiality that pursues this over-agents and ultimately threatens to break the blood-brain barrier between the application layer and the OS layer, connecting all these individual services (s) blurring their data, “concluded Whittaker.

If the message app as a signal had to integrate with AI agents, it will undermine the confidentiality of your messages, she said. The agent must access the app to send a message to his friends, as well as download the data to summarize these texts.

Her comments followed remarks, which she made earlier during the panel on how the AI ​​industry was built on a mass -collected data collection model. She said that “the greater is the better AI paradigm” -which means that the more data, the better they have potential consequences that she did not consider good.

With Agentic AI, Whittaker warned that we would further undermine confidentiality and security in the name of “Magic Genie Bot, which will take care of the urgency of life,” she concluded.