California requires training data to be disclosed
A requirement took effect to publish a description of training data: sources, quantity, types, legal status, presence of personal information and of synthetic data.
Why it matters
A developer in California must publish a high-level summary of what a model was trained on; until then that was a trade secret by default. The European Union’s duty on providers of general-purpose models, which refers to a public summary of training content, had applied since 2 August 2025.
The law covers any system available to residents of the state and requires a description for every release since 2022. The list of required items is detailed: dataset owners, the number of data points in ranges, whether the material is under copyright, whether the data was bought or licensed, whether personal data is present, what modifications were applied. The exemptions are narrow: security systems, aviation, federal national defence. Crowell & Moring note that the statute does not define what “high-level” means and specifies no enforcement procedures or penalties, so it is unclear which agency will enforce it.