xAI fails to block California data-disclosure law
AI developers with models accessible in California are subject to mandatory disclosures about training dataset sources, collection timing, licensing or purchase status, use of synthetic data, and whether training data includes personal information or IP-protected material.
Change
California enacted a law, effective January 2026, requiring AI developers whose models are accessible in the state to publicly disclose training dataset sources, collection dates and ongoing status, licensing and copyright status, and whether datasets include personal or synthetic data.
Why it matters
The law creates a legal obligation to provide dataset-level provenance and licensing details for models offered in California, limiting firms' ability to keep those details secret. Companies claiming trade-secret or constitutional protections must now identify specific, unique datasets and show concrete harms rather than relying on general assertions.
Implications
- — AI developers whose models are accessible in California must compile and publish dataset-level provenance, collection dates, licensing and copyright status, and the extent of synthetic or personal data to meet AB 2013 disclosure requirements.
- — In-house and outside counsel for AI developers must create detailed records and legal arguments that identify any datasets claimed as trade secrets and demonstrate specific, concrete harms to support exemptions in litigation.
Unlock the decision layer.
Know what's at risk and what to do next.
- Implications: What this forces you to change — operations, exposure, or compliance.
- Who is affected: Which roles, contracts, and obligations are exposed.
- What to watch: Binding deadlines and enforcement dates.
- Real-time alerts: Delivered the moment a binding change is published.
- Ask AI: Ask what this means for your specific role.
No credit card · 14-day trial · Active in seconds
Unlock the decision layer