Signal president warns AI chatbots could become surveillance tools

Whittaker’s core argument was that autonomous AI agents, device scanning and digital advertising are converging into what she described as a new architecture of surveillance. That critique lands sharply because Signal has built its identity around the opposite model: a nonprofit messaging service that uses end-to-end encryption, says it cannot read users’ messages or listen to calls, and is designed around minimal data collection. Signal Foundation was formed in 2018 to support Signal Messenger, which originated in 2012.

The warning is part of a longer campaign by Whittaker against agentic AI and the way it is sold to consumers. At SXSW in Austin, Texas, in March 2025, she argued that AI agents are marketed as convenient assistants that can look up concerts, book tickets, schedule events and message friends, but that doing so would require broad access to a user’s browser, calendar, credit cards, messaging apps and other sensitive data. She said that would amount to something like root-level permission across a device, and that powerful models would likely have to run in the cloud rather than on-device, deepening privacy and security risks. She also described the idea as “putting your brain in a jar.”

She returned to the theme at the AI for Good Global Summit in Geneva, Switzerland, on July 8, 2025, warning that agentic AI could reach a “very dangerous juncture.” The summit framed the systems as “magic genie bots” promising to manage daily life, but her argument was that the convenience comes with a steep cost in access to private data.