Guardian profile shows emotional risks of AI companion bots

While not a technical announcement, this long‑form profile is strategically important because it surfaces the human side of continuous AI cohabitation before we get to full AGI. Joanna Stern uses chatbots to answer texts, manage health information, run her household and even simulate a romantic companion—only to find that the line between “just software” and “being” blurs uncomfortably fast.([theguardian.com](https://www.theguardian.com/technology/2026/jun/04/my-year-with-robots-joanna-stern-ai)) Her conclusion—especially her plea to ban companion bots for kids—signals that even enthusiastic tech users are hitting emotional limits.

For labs racing toward more agentic, emotionally responsive models, this is a flashing yellow light. The very qualities that make advanced systems engaging and helpful also make them sticky in parasocial ways, raising mental‑health and manipulation risks. That, in turn, will attract regulatory attention: think limits on intimacy‑oriented agents, mandatory disclosures, or age‑based restrictions analogous to those now being discussed for social media.

As models approach AGI‑like capabilities, managing the social experience of interacting with them will be as important as managing raw competence. Stories like this will influence how governments, parents and educators feel about deploying highly capable systems in homes, schools and care settings, which can either accelerate or stall societal adoption curves.