Signal president Meredith Whittaker warned on Friday that agent AI could include a risk for user confidentiality.
Speaking on stage at the SXSW conference in Austin, Texas, the defender of secure communications, qualified the use of AI agents to “put your brain in a pot”, and warned that this new IT paradigm – where AI performs user tasks – has a “deep problem” with user confidentiality and security.
Whittaker explained how AI agents are marketed as a means of adding value to your life by managing various online tasks for the user. For example, AI agents could assume tasks such as concerts, ticket booking, event planning on your calendar and messaging to your friends that it is reserved.
“So we can just put our brain in a jar because the thing is to do that and we don’t have to touch it, right?”, Whittaker friends.
She then explained the type of access that the AI agent should perform these tasks, including access to our web browser and a way to drive it as well as access to our credit card information to pay the tickets, our calendar and the messaging application to send the text to your friends.
“We should be able to drive this [process] In the whole of our system with something that looks like a root authorization, accessing each of these databases – probably clear, because there is no model to make this figure, “warned Whittaker.
“And if we are talking about a sufficiently powerful model … IA which offers this, there is no way that it happens on the aircraft,” she continued. “It is almost certainly sent to a cloud server where it is treated and sent back. There is therefore a deep problem with security and privacy which haunts this media threw around agents, and which ultimately threatens to break the blood-brain barrier between the application layer and the operating system layer by joining all these distinct services [and] Coring their data, ”concluded Whittaker.
If an application of messaging as a signal should be integrated into AI agents, it would undermine the intimacy of your messages, she said. The agent must access the application to send SMS to your friends and also remove the data to summarize these texts.
Her comments followed the remarks she made earlier in the panel on the way the AI industry had been built on a monitoring model with mass data collection. She said that the “biggest is a better paradigm of AI” – which means that the more the data, the better – had potential consequences than it thought was good.
With the AI agent, Whittaker warned that we are more careful about privacy and security in the name of a “magic genius bot that will take care of the requirements of life,” she concluded.
