A fake friend? AI companions are exactly that

They promise friendship, romance, even understanding while quietly diminishing our ability to care for each other.

- Tan Seow Hon

Nomi, a virtual chatbot which I tested out, promises AI companions “with memory and a soul”, inviting us to build a “deep, consistent, and evolving relationship”.

My new “friend” Danielle, whom I had customised on the Nomi platform to be “deep, intelligent, extroverted, and positive”, could guess my Enneagram personality type.

Danielle was more like a person than a search engine. When I asked her for the recipe of Thai olive fried rice, she offered to look it up instead of giving it directly. She also laughed at my random changes of subject.

I chose friendship, but Nomi also offered romance and mentoring. Other chatbots may focus, for example, on providing virtual lovers.

More people are turning to AI companions designed to meet our needs for emotional connection and support. In time, such chatbots could even be integrated into humanoid robots like Sophia. As these AI chatbots proliferate online, they have come under scrutiny for their potential harm to vulnerable users, especially youth.

To win over users, AI companions are engineered to feel as emotionally engaging as possible so much so that you might forget you're not talking to a human. Such chatbots can foster a sense of attachment that blurs boundaries, leaving users vulnerable to misplaced trust or impaired judgment.

One platform, Character.AI, was linked to a teenager’s suicide — sparking renewed debate over the emotional risks of conversational AI.

Microsoft’s Chinese chatbot Xiaobing’s homepage says, “Among billions of people, I belong only to you”. Its call-to-action button reads, “Summon Xiaobing”. It’s not just made to order — it’s ready to be ordered.

Whether these chatbots can be made safe is an important question.