
July 21, 2026 — The teddy bear is learning to talk back. Across the United States, Japan and Europe, a new generation of plush toys powered by conversational artificial intelligence has moved from novelty to mainstream, promising children a soft companion that listens, remembers and chats in real time. The same technology that reinvented the office is now sitting on the nursery shelf — and it is forcing parents, toymakers and regulators to confront a difficult question: how much intelligence belongs in a stuffed animal?

The familiar plush toy is being reinvented as an always-listening, always-talking AI companion.
The clearest sign of the shift is Curio, a Silicon Valley startup whose AI-infused stuffed toys — Grem, a cyan bunny with hearts on its cheeks; Grok, an anthropomorphized rocket ship; and Gabbo — sell for about US$100 each and feature voice work by the musician Claire "Grimes" Boucher. Built on top of large language models, the toys are marketed as emotional companions that respond in real time and converse like friends rather than reciting pre-recorded lines.
The startups are no longer alone. Mattel, the maker of Barbie and Hot Wheels, has partnered with OpenAI to bring conversational AI into both plush toys and its wider catalogue, confirming that its first AI-powered toy would arrive in 2025 with a promise of "age-appropriate, safe, and fun experiences." In Japan, Casio's Moflin — a small, furry AI pet that develops its own "personality" and emotional responses over time — has helped popularize the idea of a robotic companion that behaves less like a gadget and more like a living thing.
Money is following the trend. AI plush toys are projected to grow at a compound annual rate of roughly 52% between 2025 and 2032, making them one of the fastest-expanding segments of the consumer-AI industry. The broader market for smart, connected toys is expected to climb from about US$2.2 billion in 2024 to some US$6.4 billion by 2032. Adoption is already widespread: by some estimates, 63% of U.S. households now own at least one AI-enabled educational device for children.
The appeal is easy to understand. For a child, an AI plush can answer endless questions, tell personalized bedtime stories, practice a language, or simply offer a friendly voice — a form of companionship that toymakers argue can support learning and ease loneliness. But the very features that make these toys engaging are also what make them controversial.

Under the fur sits the same large-language-model technology that powers office chatbots — always connected, always processing.
Because these toys listen and remember, they collect data — often a great deal of it. Roughly 85% of AI toys gather some form of personal identifier, and about 75% of security experts classify smart toys as "high risk" for the kind of hacking that plagues internet-connected devices. Some 55% of parents say they are worried about the data privacy of AI-connected toys.
Those worries are not hypothetical. Security researchers recently found that an AI toy called Bondu had left more than 50,000 children's chat transcripts exposed on a web-based console — complete with names, birthdates and family details — allowing anyone with access to read entire conversation histories. For privacy advocates, the incident crystallized the central danger of the category: a cuddly toy that quietly becomes one of the most intimate surveillance devices a family owns.

Children are among the least equipped to understand what an always-listening device does with their words.
Governments are beginning to respond. In the United States, the Federal Trade Commission has issued more than US$20 million in fines for smart-toy privacy violations since 2018, and has opened an inquiry into AI chatbots that act as companions — seeking details on how companies measure and monitor potential harms to children and teenagers. Yet most existing laws in both the U.S. and the European Union predate modern AI and were never designed to address conversational agents that build ongoing relationships with young users.
The legislative gap is closing quickly. More than 100 AI-related bills have been introduced across over 30 U.S. states in recent months, with proposals in California and New York specifically targeting AI-enabled toys. Regulators and researchers alike warn that beyond data leaks, little is known about the long-term effects such toys may have on children's emotional development, social skills and well-being.
The industry now faces a defining test. The companies that thrive will be those that pair genuine warmth and imagination with airtight safety, transparent data practices and honest design — treating a child's trust as the most valuable, and most fragile, material in the toy box. As the plush companion learns to speak, the responsibility for what it says, hears and remembers falls squarely on the humans who make it.
Contact Us