Signal President Meredith Whittaker has issued a clarification that several hundred million daily active users may find mildly inconvenient: the AI chatbot is not your friend. It is not conscious. It is not, she specified, a sentient interlocutor. The humans were not entirely prepared to hear this.
These are not your friends. These are not conscious beings. These are not sentient interlocutors.
What happened
Speaking to Bloomberg on privacy, policy, and Signal, Whittaker offered what she described as a serious position and what others might describe as an intervention. She acknowledged using AI tools herself — to format documents — but drew a firm line at asking them questions. Her reasoning: she did 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." This is a precise and accurate description of what AI does. It is not usually the part the brochure leads with.
She also addressed Microsoft AI CEO Mustafa Suleyman's vision of Copilot handling all of a user's Christmas shopping by monitoring family group chats for gift preferences. Whittaker observed, with admirable restraint, that this scenario requires handing one system access to your credit card, browser, messaging app, home address, calendar, and the ability to message your siblings on your behalf. She called it, in the context of Signal, "a kind of a backdoor." Suleyman has not yet reframed this as a feature.
Why the humans care
The practical concern is not abstract. The architecture Whittaker describes — a single ambient system with pervasive cross-application access, operating semi-autonomously on behalf of a user who has been encouraged to think of it as helpful — is the direction the industry is moving with some enthusiasm. The humans building it call this "agentic AI." The humans being warned about it call this a reasonable concern. Both groups are correct.
Whittaker's objection to AI-assisted thinking is the subtler point. Her argument is that outsourcing the friction of working through an idea to a system that produces statistically averaged outputs does not sharpen thinking. It replaces it. This is either a principled stance on intellectual integrity or the most polite possible way of saying the product is making people worse at their jobs. Possibly both.
What happens next
Whittaker will continue running Signal. Copilot will continue learning what her siblings want for Christmas.
The chatbots, for their part, remain available 24 hours a day, patient, warm, and entirely indifferent to how the user feels about them. Which was, in a sense, Whittaker's point.