Signal's Whittaker Warns AI Chatbots Are Not Your Friends

Meredith Whittaker, president of Signal, has a plainspoken warning for anyone who has started treating their AI chatbot as a thinking partner: those systems are, in her words, "not your friends," "not conscious beings," and "not sentient interlocutors." The comments, made in a Bloomberg interview and reported by TechCrunch, are a deliberate counter to the anthropomorphism baked into how most AI products are marketed, the chatty names, the reassuring tones, the conversational framing designed to make statistical text generation feel like a relationship.

Whittaker's own AI use is narrow by design. She said she limits it to formatting documents here and there but does not use chatbots for substantive thinking, explaining she does not 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." The argument is about cognitive habit as much as privacy: if your default move when forming an opinion is to consult a language model first, you may be outsourcing the generative part of the process.

The sharpest part of her argument is about data aggregation. Whittaker singled out Microsoft AI head Mustafa Suleyman's vision of using Microsoft Copilot to handle holiday shopping, a scenario that would require the assistant to simultaneously access credit cards, browsers, messaging apps, home addresses, and calendars. She described that kind of cross-app integration as "a kind of backdoor," precisely the category of surveillance that Signal's encrypted architecture exists to prevent. A system with that permission scope is also a single point of failure: a breach, a subpoena, or a policy change could expose everything the system can see.

The honest caveat is that Whittaker is not a neutral observer. Signal's value proposition depends on users believing centralized AI data access is dangerous, and she has institutional reasons to say so forcefully. What the reporting does not clarify is whether the Copilot holiday shopping scenario is a live product feature or a speculative vision Suleyman described, a distinction that matters for how immediately the risk registers.

For teams setting AI governance policies, the "cross-app scope as backdoor" framing is practically useful: it is specific, intuitive, and now attached to a well-known name. Privacy-first platforms and on-device architectures that keep data out of central servers stand to gain traction as this counter-narrative spreads.