Signal President Warns Against Treating Chatbots as Friends

Meredith Whittaker, president of Signal, spoke publicly about interpersonal expectations and privacy risks around AI chatbots. India Today reports she told Bloomberg, "These are not your friends. These are not conscious beings. These are not sentient interlocutors." Yahoo reports she said she uses AI "to format a document here and there" but added, "I don't ask them questions. I'm very serious about my thinking and writing, and I don't want the process of working through an idea ... to be foreclosed or eclipsed by the response of a system that's averaging what's already out there." Yahoo also quotes her describing a hypothetical Copilot scenario with broad access across services as "a kind of a backdoor."

Industry observers routinely note that modern chatbots are statistical models trained on large datasets and do not possess consciousness or intent. This technical characteristic explains why experts repeatedly caution against anthropomorphizing systems: conversational fluency can create an illusion of understanding even though responses are generated by pattern-matching and probability. For practitioners, that gap between apparent empathy and underlying mechanism affects how conversational agents should be evaluated for safety, disclosure, and human-in-the-loop workflows.

Reporting places Whittaker's comments alongside broader public debate about privacy, integration, and user trust. India Today and Yahoo situate her remarks amid recent episodes where vendors adjusted or removed personas and where companies proposed deep integrations that would link assistant capabilities to personal data across services. Observers following privacy and product design note persistent tensions between convenience-driven integrations and the risk surface those integrations introduce when an assistant gains cross-application access.

Companies building or deploying conversational agents should expect user expectations to be shaped by design choices that make models appear personable. Industry-pattern observations show that transparency about model limitations, careful access controls, and explicit user consent flows matter more when assistants are framed as agents acting on users' behalf rather than as drafting aids.

Watch for provider disclosures or defaults about assistant access to messages, payments, browser data, and calendars; regulatory attention on cross-app data sharing; and product design signals (UI affordances, consent dialogs) that either encourage or discourage anthropomorphic interactions.