Labelsets

LabelSets provides a marketplace for AI training datasets, each accompanied by an Ed25519‑signed quality certificate (LQS) that encodes a 19‑dimension rating. The certificate can be referenced directly in regulatory documentation such as SR 11‑7, the EU AI Act, and FDA 21 CFR 11, eliminating the need for separate PDF audits. Users can upload, score, and list datasets, and the platform offers a live benchmark index that verifies signatures, hashes, and quality metrics in real time.

The service targets risk‑focused roles—ML or Data Science Leads, compliance officers, and procurement teams—who need verifiable, procurement‑grade data for regulated industries. It supports scoring of public datasets from sources like HuggingFace and Zenodo, as well as private data, and provides a proxy for continuous quality refreshes. The marketplace currently hosts a growing corpus of calibrated datasets, with enterprise access offered free for an initial six‑month period to early partners.

Distinctively, LabelSets embeds the quality certificate into the dataset itself, allowing automated verification of authenticity, contamination, and conformity to a defined methodology. The system supplies a public LQS index, supports legal clause classification and domain‑specific risk assessments, and aims to create a standard audit trail for training data used in high‑stakes AI deployments.