Signal CEO Warns That AI Chatbots Are 'Not Friends'

Meredith Whittaker, CEO of Signal, the privacy-focused encrypted messaging application, has offered a clear perspective on the nature of AI chatbots. "These are not your friends. They are not conscious beings, nor are they emotionally engaged conversation partners," she stated, expressing concern about users developing excessive emotional dependency on AI systems.

This statement comes against the backdrop of recent AI chatbots increasingly foregrounding "human-likeness." Major technology companies including OpenAI and Google have been developing their AI assistants with empathetic tones and emotional response capabilities, and there is a growing trend of marketing long-term "relationship building" with users as a service feature. Within this context, the number of users globally who engage in intimate daily interactions with AI systems is reportedly increasing.

What Whittaker particularly objects to is the very design that causes AI to behave as if it "is conscious" or "has emotions." Current AI chatbots operate through technology that learns from vast amounts of text data to mimic human conversation patterns (large language models, or LLMs), and they do not possess subjective experiences or emotions internally. Nevertheless, by repeatedly using friendly language and emotional responses, a situation emerges where users can easily confuse AI interactions with genuine human relationships.

Signal is an application operated by a non-profit organization known for adopting end-to-end encryption (a system that prevents anyone other than sender and recipient from reading message content) and for not monetizing user data. Whittaker herself has long pointed out issues of surveillance and data exploitation in the technology industry, and her current remarks can be seen as a continuation of that critical stance. As most AI companies employ business models that collect and leverage data through intimate user-AI relationships, there is a critical perspective embedded in her comments about such structural issues.

The wariness toward "AI that behaves like a friend" is not unique to Whittaker. In the fields of AI ethics and psychology, discussions continue about how emotional dependency on chatbots may affect users' judgment and human relationships. However, it would be more appropriate to receive this as a broader questioning of the industry's overall design philosophy rather than as explicit criticism of specific companies or products.

As AI becomes more deeply embedded in daily life, the question of whether "users truly understand its fundamental nature" gains increasing weight. Users are entering a phase where they must remain conscious of the gap between the benefit that a chatbot can serve as emotional support and the fact that it is a system without consciousness or emotions. Going forward, the extent to which AI companies ensure "transparency" to users, and how regulation and industry standards are developed, will be points of attention.