California’s AI Training Data Transparency Act Draws Trade Secret Conflict
California’s AB 2013, the Generative Artificial Intelligence Training Data Transparency Act, is creating a legal collision between state disclosure mandates and federal trade secrets protections, Reuters reported on May 18. The law requires AI developers to disclose the datasets used to train generative models.
That requirement conflicts directly with companies’ ability to protect proprietary training pipelines as trade secrets.
The Core Conflict
The law, known as the TDTA, compels AI developers to publish summaries of the data sources used to train their systems. Affected companies argue that revealing training data details crosses into territory protected under the federal Defend Trade Secrets Act.
A Reuters analysis published May 18 examined how practitioners are navigating the gap between the two legal frameworks. The tension is particularly acute for foundation model developers whose competitive advantage depends on proprietary curation of large datasets.
Also Read: MEXC Launches Free AI Trading Strategy Tool for All Users
Background
California signed AB 2013 into law in 2024, making it one of the first U.S. states to impose affirmative disclosure obligations on generative AI developers.
The law targets models trained on data collected or processed after January 1, 2022. Under the statute, developers must post training data summaries on their websites, covering the general categories of data and whether copyrighted material was included.
The TDTA does not require companies to hand over raw datasets. The law’s authors intended it to improve public accountability for AI systems without forcing full data disclosure, but legal practitioners say the line between permissible summary and protectable trade secret is unclear in practice.
Also Read: Vitalik Buterin Says AI Formal Verification Could Strengthen Crypto Security
Industry Response
Technology industry groups have pushed back on the TDTA since its passage.
Critics argue that even summary-level data disclosure can enable competitors to reverse-engineer model architectures or identify proprietary curation strategies. Practitioners advising AI companies are now recommending that clients seek legal opinions on whether specific disclosures qualify for trade secret protection before publishing compliance documentation.
Courts have not yet ruled directly on the conflict between the TDTA and federal trade secrets law, leaving companies in an uncertain compliance posture.
What to Watch
A legal challenge to the TDTA’s disclosure requirements is widely anticipated. If a court finds that state-mandated disclosure strips trade secret protection, the ruling could reshape how California regulates AI transparency.
Federal preemption arguments are also in play. The outcome would affect not only California-based developers but any company training generative AI models on data touching California residents.
Read Next: OriginTrail Surges 33% as AI Knowledge Infrastructure Narrative Gains Traction
