Signal’s Meredith Whittaker: AI Chatbots Are Not Your Friends
Signal president Meredith Whittaker warned users that AI chatbots are not conscious or sentient. Here's what that means for how you use these tools.
LUMIEN4 min read
Signal president Meredith Whittaker issued a blunt public warning about how people relate to AI chatbots, stating plainly that they are "not your friends," "not conscious beings," and "not sentient interlocutors." The comment, reported by TechCrunch on June 20, 2026, cuts against a wave of AI products deliberately designed to feel personal and emotionally present. Whittaker's position carries weight given Signal's long-standing focus on user privacy and her own prominence as a vocal tech critic.
What happened
Meredith Whittaker, president of the Signal Foundation, made a pointed statement about the nature of AI chatbots. Her exact words, as reported by TechCrunch: “These are not your friends. These are not conscious beings. These are not sentient interlocutors.”
The statement is a direct challenge to how many AI products are currently positioned. Companion apps, emotionally responsive assistants, and “personalized” chat products increasingly encourage users to form ongoing, personal relationships with software. Whittaker wants people to resist that framing.
Why it matters
This is not a fringe concern. A growing number of consumer AI products are built around emotional engagement. The design goal is often to make the interaction feel warm, continuous, and personal. That serves the business model. It does not necessarily serve the user.
When someone treats a chatbot as a confidant or companion, a few things tend to follow:
- They share more personal information than they would with a clearly transactional tool.
- They extend trust that has not been earned and cannot be reciprocated.
- They may rely on outputs for emotional support or decision-making in ways the technology is not equipped to handle responsibly.
Whittaker’s background makes her framing specific. Signal is built on the premise that users deserve honest tools, not ones that manipulate behavior for engagement. Her warning fits that philosophy directly.
There is also a broader structural point here. AI chatbots are products built by companies with commercial interests. The “personality” of a chatbot is a design choice made by engineers and product teams, not evidence of inner experience. Conflating the two is a mistake with real consequences.
Our take
Whittaker is right, and the point is worth repeating clearly for anyone using these tools in a business context.
AI assistants are useful. We use them. Our clients use them. They save time on writing, research, summarizing, and pattern-spotting. None of that requires them to be sentient, and none of it becomes more reliable because the interface feels friendly.
The risk for business owners is slightly different from the consumer companion risk Whittaker is likely addressing. It is less about emotional attachment and more about misplaced confidence. A chatbot that sounds confident and warm can still be wrong. A tool that remembers your preferences is not the same as a tool that has your interests at heart.
The anthropomorphism baked into many AI products is a design choice, not a feature you asked for. Being aware of that helps you use the tools more clearly: check outputs, do not share sensitive data you would not hand to a vendor, and do not mistake fluency for accuracy.
Whittaker’s framing is blunt because blunt is what is needed when the products themselves are working in the opposite direction.
What to do about it
A few practical habits worth building into how your team uses AI tools:
- Treat chatbot output as a draft, not a verdict. Review it the same way you would review work from a contractor you just hired.
- Set a data policy. Decide clearly what information is acceptable to paste into a chatbot and what is not. Customer data, financial details, and unreleased product plans are common areas where teams overshare.
- Notice when a tool is designed to feel personal. That is a product decision, not a technical property. Ask what interest it serves.
- Separate usefulness from trustworthiness. A tool can be genuinely useful and still have no stake in your outcomes.
The most productive way to use AI is as a fast, capable, sometimes-wrong piece of software. That framing keeps you in the driver’s seat.
