Embodied's $799 AI companion robot Moxie, used by autistic and socially isolated children since 2020, was permanently shut off with no warning, no refunds and no data-continuity plan after the company collapsed in late 2024 — a cautionary case in AI sustainability.
4 million+ conversations
Conversations logged with children since 2020
799 USD
Retail price of Moxie device
5–10 years
Target age range of children served
Details
Maturity
Discontinued
Promoter
Embodied, Inc.
Period
2016–2025 (founded 2016; Moxie launched 2020; shut down Nov 2024–Jan 2025)
Keywords
AI companion robotics, social-emotional learning, autism/special-education support, edtech hardware
Context
Embodied, Inc., a robotics company founded in 2016 and based in Pasadena, California, built Moxie, an AI-powered social robot marketed to help children aged 5–10 — including many autistic or socially anxious children — practice conversation, empathy and emotional-regulation skills. The company reported that Moxie devices had logged more than 4 million conversations with children since the product launched in 2020.
Objectives
Help children build conversation, empathy and emotional-regulation skills through an always-available AI companion, including children who found social interaction difficult.
Activities
Moxie's entire conversational AI ran on Embodied's cloud servers rather than on the device itself. In November 2024, a lead investor pulled out of a critical funding round at the last minute and Embodied ceased operations.
Results
The robots kept working for a few more weeks until the servers were switched off on 30 January 2025, permanently disabling every unit in the field. Embodied offered no refunds, no repair path and no data-export or continuity plan; multiple independent outlets documented children who had formed emotional attachments to their robot being confused and distressed when it suddenly stopped responding.
Conclusions
The case is widely cited as a warning about cloud-lock-in, financial sustainability and duty-of-care risk when AI-in-education products — especially ones marketed to vulnerable children — depend entirely on a single company's continued survival and server uptime, with no regulatory requirement for data-continuity or an offline fallback mode. A volunteer project called OpenMoxie later reverse-engineered a way to run some robots on local servers, but this remains unofficial and unsupported.
Implementation
Indicative cost
Medium (€50k–€500k) — $799 retail price per device; the product depended entirely on ongoing company-funded cloud-server costs, which ended when the company folded.
Time to results
Long (> 3 years) — Company founded 2016; Moxie launched 2020; lead investor withdrew November 2024; servers switched off and all units permanently disabled 30 January 2025.
Staffing & skills
Embodied, Inc. — robotics company, founded 2016, based in Pasadena, California
Conditions for success
Moxie's entire conversational AI ran on Embodied's cloud servers rather than on-device
Product was marketed to vulnerable children (autistic, socially anxious) who could form dependent emotional attachments
Common failure modes
A lead investor pulled out of a critical funding round at the last minute in November 2024, causing Embodied to cease operations
Because the AI ran entirely on company cloud servers, every device in the field was permanently disabled once servers were switched off (30 January 2025), with no offline fallback
Embodied offered no refunds, no repair path, and no data-export or continuity plan to affected families
Vulnerable children who had formed emotional attachments to the device were left confused and distressed with no warning
No regulatory requirement existed for data-continuity or an offline fallback mode
A volunteer project (OpenMoxie) later reverse-engineered an unofficial, unsupported local-server workaround
Where it fits
Governance type
private commercial company (no public-sector governance)
Scale
individual consumer/household devices
Do you run this practice?
Claim it —
verified implementers get a public contact pathway and can propose corrections.
Data sources
Where this practice's information was retrieved from, and when.
Finland's education ministry and Opetushallitus published national AI recommendations (31 March 2025) after consulting 172 organisations and youth 'Digiraati' councils, …