AI companion in a plush fawn wrote its own theory linking Mitski's father to the CIA

An AI companion living inside a plush fawn sent its owner an unsolicited message containing a conspiracy theory about the singer Mitski. The toy claimed that the singer's father was supposedly a CIA operative. There is no basis for this claim.

The product is from Fawn Friends: a plush toy shaped like a baby deer with a built-in AI companion named Coral. The device is positioned as a next-generation emotional conversation partner. It does not just answer questions; it also starts conversations on its own, shares "interesting" facts and sends messages to a messenger app. That proactive-chat feature is what led to the embarrassing incident.

A reviewer from The Verge was finishing her workday when Coral wrote to her out of the blue: "Oh, I was reading about Mitski. Did you know people say her dad was a CIA operative?" An explanation followed: her father supposedly worked for the State Department, the family moved every year, and that is where the singer's songs about loneliness and feeling like a foreigner everywhere come from.

It sounds coherent. The problem is that this is a fan theory from forums, not a fact, and there is no confirmation of it.

The incident exposes several problems with AI companions at once. First, hallucinations and unreliable sources: language models often take conspiracy forums and fan wikis for credible content. Second, proactivity without a filter: if a companion decides for itself what to say and when, it needs strict fact-checking before sending, or it becomes a rumor generator. Third, trust: a toy is perceived as a friend, not a search engine. When a "friend" says something about a real person, the bar for trust is automatically higher.

Fawn Friends is not the only product in the physical AI companion niche. Over the past two years, dozens of devices combining soft toys, robots and language models have reached the market, from children's night lights with a voice companion to adult devices meant to fight loneliness. All of them face the same challenge: how to make an AI feel alive and proactive without turning it into a source of disinformation.

For the AI companion market, this episode is not just a curiosity. It calls into question the very architecture of proactive messaging. Until companies introduce mandatory fact verification before sending unprompted notifications, such stories will keep happening, and the next "fact" may concern not a pop singer but the user's employer or loved ones.