Huge wearable dataset announced

A wearable is a body sensor that turns sleep, heart rate, steps and recovery into a time series — a stream of numbers collected day after day. ORO AI says it has assembled one of those streams at unusual scale by pooling data from devices made by Whoop, Fitbit, Oura and Garmin. (getoro.xyz) (x.com) ORO described the release in an April 2026 X post as a wearable dataset with more than 600,000 contributors and more than 11 million monthly data points, and said artificial intelligence labs are already using it. The company’s website says people contribute data by connecting existing accounts and completing in-app “quests.” (x.com) (getoro.xyz 1) (getoro.xyz 2) The company, based in New York, raised a $6 million seed round announced on April 16, 2025, led by a16z crypto CSX and Delphi Ventures. In that announcement, ORO said it was building a “multimodal” contribution app, meaning one system that can combine different kinds of inputs, including connected-account data and new tasks completed inside the app. (getoro.xyz) That pitch lands as health artificial intelligence developers hunt for data outside the public internet. ORO says “publicly available data is running dry” and argues that private device data has become the next source for model training. (getoro.xyz 1) (getoro.xyz 2) Wearable data is attractive because it is continuous, personal and tied to physical signals rather than typed text. A 2025 review in *Frontiers in Digital Health* said wearable-AI systems can support continuous monitoring and prediction, while also raising concerns about informed consent, bias, privacy and opaque decision-making. (frontiersin.org) ORO says its system does not expose raw data directly to outside model builders. In product posts and its FAQ, the company says it uses Trusted Execution Environments — secure hardware enclaves — plus cryptographic tools including zkTLS and, in some cases, Multi-Party Computation to process data without showing the underlying records to people or node operators. (getoro.xyz 1) (getoro.xyz 2) The open question is provenance: exactly what contributors agreed to, what downstream labs can do with model outputs, and how those permissions map back to each source platform. ORO says users “control what you share” and are “fairly compensated,” but the company’s public materials reviewed here do not spell out dataset-level terms for outside researchers or labs. (getoro.xyz) (getoro.xyz) The source platforms already have their own rules and data pathways. Oura and Whoop both offer developer access for authenticated user data, Fitbit’s privacy policy says collected information may be used to develop services that “may include generative artificial intelligence models,” and Garmin markets Garmin Health for research and clinical trials. (cloud.ouraring.com) (developer.whoop.com) (fitbit.com) (garmin.com) What ORO has put on the table is not a new wristband or ring, but a new market for the exhaust those devices already produce. The next test is whether large health-data pools can scale faster than the questions attached to consent, auditability and who gets paid when a model learns from your body. (getoro.xyz) (frontiersin.org)