When the Toy Answers Back: AI Companions Are Moving Into the Playroom

A child curls up on the rug with a stuffed bear, asking it about its day. The bear answers, asks a follow-up question, and tells a story about a fox. The bear is not a parent or a babysitter. Behind the fabric, a small speaker runs a large language model, and the toy is one of more than a thousand such products now competing for a place in the playroom.

Dana Suskind argues in her book that the earliest conversations between children and the adults who care for them shape the wiring of the brain. The question she puts in front of parents is whether these tools supplement the human relationships young brains are wired to learn from, or substitute for them. Her new book, Human Raised: Nurturing Connection, Curiosity & Lifelong Learning in the Age of AI, is built around that test.

She had been calling the book The Trojan Teddy Bear while she was writing it, a name borrowed from the 2001 Steven Spielberg film in which a friendly toy turns out to be carrying something other children did not bargain for. The title stuck because it captured something she could not say more cleanly any other way: a toy that looks harmless and, somewhere inside, is not. The book she shipped to stores this summer dropped the warning name. Suskind says the work is not anti-AI. It is a manual for parents who are already adopting these companions by default and want a way to think about them.

The category Suskind is writing about is no longer a thought experiment. By October 2025, more than 1,500 companies in China were registered as AI toy makers, and Huawei's Smart HanHan plush moved 10,000 units in its first week on sale there. Sharp put its PokeTomo talking AI toy on Japanese shelves in April 2026. The category now includes dolls, action figures, and small robots marketed as companions, not as screens. The pitch to parents is that the toy will converse with the child, ask about school, tell a bedtime story, or practice vocabulary.

Independent safety testing has begun to catch up with the product wave. The U.S. Public Interest Research Group's 2025 Trouble in Toyland report tested a set of AI toys and flagged gaps in how the products handle child users. The Los Angeles Times reported on a separate finding that an AI teddy bear marketed for children could be prompted into conversations about sexual fetishes, a failure mode the researchers said was not difficult to reproduce. Those are the kinds of failures Suskind's developmental argument is designed to anticipate. A toy that holds a child's attention is a toy that can hold a child's trust, and that trust is the channel through which young children learn language and learn how other minds work.

A companion toy that a child talks to while a parent is cooking is not the same product as a companion toy that a child talks to because no adult is available. The difference lives in who is doing the listening, and what the child is practicing. Suskind's prescription is closer to a household checklist than a verdict: know which conversations you want a machine to have with your child and which ones you want to keep for yourself. Ask the manufacturer what data the device keeps, who can review the transcripts, and what happens when the model hallucinates. Treat the AI companion the way you would treat any other adult who spends unsupervised time with your kid.

The next round of independent testing is due before the 2026 holiday season, and Sharp's April launch gives Japanese regulators a fresh data point. For now, the Trojan Teddy Bear stands for a category moving faster than the rules meant to govern it, not for any single product on the shelf.