California AB 2013: Generative AI Training Data Transparency Act

Training data transparency

Must AI developers disclose what they trained on?

A summary or on request

Developers must post a high-level summary of the data used to train generative AI

Holding

Developers of generative AI systems made available to Californians, released on or after 1 January 2022, must publish on their website a high-level summary of their training datasets, including sources, volume, collection methods and whether the data includes copyrighted material or personal information. The law takes effect on 1 January 2026.

The first US training-data disclosure law; it has no separate enforcement mechanism and does not define 'high-level'.

Sources