Binding law — in force
GenAI developers must publish a training-data summary
Developers of generative AI made publicly available to Californians (released/substantially-modified on or after 2022-01-01) must post on their website a high-level summary of the training datasets covering 12 enumerated items: sources/owners, purpose, data-point counts, data types, copyright/trademark/patent status, purchased/licensed status, whether personal information or aggregate consumer information is included, cleaning/processing, collection time period, first-use dates, and synthetic-data use. Detect a generative AI codebase with no training-data-disclosure artifact.
ca-ab2013.training-data-disclosure · Cal. Civ. Code 3111(a) · official source · jurisdictions: US-CA