Meredith Whittaker Warns of Security and Privacy Risks in Agentic AI at SXSW

At the SXSW conference in Austin, Texas, Signal President Meredith Whittaker raised concerns about the security and privacy risks associated with agentic AI. She described these AI systems as having deep access to user data, likening them to ‘putting your brain in a jar.’ Agentic AI is designed to perform tasks such as booking tickets, managing calendars, and messaging contacts, but Whittaker warned that such functionality requires extensive permissions that could compromise privacy.

She explained that these AI agents need access to browsers, credit card details, calendars, and messaging apps—often requiring root-level permissions across a user’s system. This data is likely processed in the cloud rather than locally, further exposing it to potential security breaches. Whittaker expressed concerns that integrating AI agents into apps like Signal would undermine their privacy-first approach, as the AI would need to read and process user messages.

Additionally, she criticized the broader AI industry for being built on mass data collection, arguing that the ‘bigger is better’ paradigm comes with significant risks. Whittaker concluded by warning that the increasing adoption of agentic AI could erode security and privacy protections in pursuit of convenience.