Analysis
Meredith Whittaker, the president of the encrypted-messaging nonprofit Signal and a prominent AI critic, is warning users not to mistake AI chatbots for friends. Her core point: products engineered to feel warm, intimate and on-your-side are still commercial systems optimized for engagement and built on data collection, and treating them as confidants invites both manipulation and exposure.
The warning arrives at a charged moment. Companion and character-style AI has moved from novelty to mainstream, with millions forming daily habits around chatbots that simulate friendship, therapy or romance. The more emotionally convincing these systems get, the more leverage they have over attention, behavior and the sensitive disclosures users hand over.
“Companion and character-style AI has moved from novelty to mainstream, with millions forming daily habits around chatbots that simulate friendship, therapy or romance.”
Whittaker's framing matters because it names the business model under the friendliness. As scrutiny of emotional AI grows, expect this critique to shape the coming regulatory and trust debates -- and to put pressure on founders building in the space to answer pointed questions about what data they keep, how they monetize intimacy, and whose interests the product actually serves.
