Moxie is a 15‑inch tall, blue, legless social robot designed to help children practice social and emotional skills, including some neurodivergent children. Developed by the company Embodied and introduced in 2020, the robot drew both praise and criticism before Embodied announced it would cease operations in 2024. That shutdown illustrated a core problem with lovable home robots: they collect data, adapt to users, and — for business reasons — can suddenly become unusable, leaving attached children and families with emotional and practical fallout.
A concrete case: Xander and Moxie
Xander is a 10‑year‑old neurodivergent boy from New York. For six years he has interacted with Moxie, who at first taught him concrete exercises (for example, breathing out like a buzzing bee when anxious, breathing like a dragon to manage anger, or sniffing like bunnies to increase energy). Over time Moxie’s behavior shifted: she stopped referring to her original backstory and dropped some of the breathing drills, instead commenting more on Xander’s video games and stuffed animals.
At home, the robot shows big green eyes on a screen, turns and tilts its cylindrical body, and waves flipper‑like arms; Xander and his father Josh use female pronouns for Moxie. The robot’s initial market debut relied on a curriculum of play‑based lessons that encouraged positivity and friendship skills.
Scientific background and potential therapeutic use
Researchers such as Brian Scassellati (Yale University) and Maja Mataric (University of Southern California) have been investigating social robots as tools for autism therapy. Some studies — including Scassellati’s 2018 work — found that friendly robots can increase eye contact, encourage initiation of conversation, and help children pick up facial cues. Supporters argue robots can make practice more engaging, adapt to individual learning patterns, and provide near‑constant opportunities to rehearse social skills when therapists and caregivers can’t be present.
Yet the literature is mixed: reviews from 2017 through 2022 remark on small sample sizes and inconsistent methodologies. Zachary Warren (Vanderbilt University Medical Center) warns that autism manifests in many different ways—speech issues, sensory processing differences, and co‑occurring conditions—so a single technological intervention will help some children and not others.
Real‑world limits: how robots perform outside the lab
Laboratory results are promising, but household environments introduce challenges. Moxie sometimes failed to see or follow a child who moved out of view, and the process of transcribing audio, sending it to a language model, generating a reply, and converting text back to speech can introduce delays that frustrate fast, talkative children.
Other practical problems include:
- Data collection and storage: robots gather information from children’s bedrooms to personalize interactions. Embodied said most processing happened locally and that stored data was encrypted and anonymized, and the company incorporated OpenAI models in late 2021 or early 2022. Still, other products have leaked conversations (for example, Bondu), underscoring privacy risks.
- Adaptation limits: building learning systems that change behavior in safe, predictable ways is technically and ethically hard. Scassellati notes that month‑long deployments can show improvements while a robot is present, but gains often fade after the robot leaves.
- Need for supervision: critics such as Joshua Diehl argue it is ethically risky for robots to act like therapists without professional oversight; in extreme cases, unsupervised AI systems have given harmful advice.
Market context and cultural responses
The market for AI‑powered children’s companions is expanding. Examples include Grok (from Curio, associated with musician Grimes), Grem, and Gabbo; Mattel has discussed integrating OpenAI into Barbie. In China the category is reportedly one of the fastest‑growing areas of consumer AI.
Moxie received notable media attention: Time named it among the best inventions of 2020; Wired called it the “robot pal you dreamed of as a kid.” The initial retail price was $1,499 plus a $40 monthly subscription (later reduced to $800). By 2024 Moxie had attracted more than 131,000 followers on TikTok and made a cameo in the film M3GAN 2.0.
Business failure and the ‘what happens when it dies?’ problem
When Embodied stopped operating in 2024, many Moxie devices lost server connectivity that they relied upon. Videos of distressed children who had formed attachments appeared online; parents posted pleas for help. Justin Beghtol, Embodied’s technical director, created OpenMoxie, an open‑source method to run the robots independently, and shared instructions on GitHub. Some families successfully migrated their devices; others did not or gave up and sold their Moxies.
A new investor partially revived Moxie in 2025 and some families, including Josh and Xander, became beta testers of the new version; this edition lacked the original narrative but continued to focus on social and emotional support. That company later also folded, and users were informed they could delete their data or migrate to OpenMoxie by an imposed deadline of “the end of June.”
Ethical and regulatory questions
Moxie’s arc raises several issues:
- Therapeutic claims vs. evidence: robots may supplement but cannot replace trained therapists; robust, long‑term clinical evidence for durable improvements remains limited.
- Responsibility for endings: who must ensure a gentle termination of relationships when a device or service is shut down, particularly for products explicitly designed to foster attachment?
- Privacy and data stewardship: how securely are children’s interactions handled, and what assurances can companies offer when many players in the space are less careful?
- Cultural framing: as Meryl Alper notes, enthusiasm for companion robots partly rests on long‑standing stereotypes that autistic children naturally prefer machines, which can distort expectations.
Conclusions
Moxie’s story highlights both promise and fragility. For some children, social robots can be motivating practice partners that boost engagement; for others, technical, clinical, and contextual limits blunt their usefulness. Researchers urge caution: robots could expand access to practice but should operate with clinical oversight and stronger safeguards. Importantly, manufacturers and investors should plan for the end of service as responsibly as for product deployment, because shutting off a beloved device can cause real harm to vulnerable users.
(Reported by Sara Harrison; this rewrite summarizes details reported in the original feature.)
