A7 Labs builds an identity registry for AI training data. Before de-identification, it resolves every person and organization in an unstructured enterprise corpus (email, Slack, contracts, CRM records) into one canonical identity, mapping aliases, identifiers, attributes and relationships so context survives tokenization. Data labs, dataset builders, de-identification vendors and enterprises training open-weight models use it to keep relationships intact in training data. Founded by the team that built the AI privacy product at an enterprise data privacy infrastructure company. Suggested category: Data De-identification and Pseudonymity Software.