xAI Challenges California’s Training Data Transparency Act

On December 29, 2025, xAI, the developer of the artificial intelligence (AI) chatbot Grok, filed a lawsuit seeking to invalidate California’s Generative Artificial Intelligence: Training Data Transparency Act (TDTA). The TDTA, which took effect on January 1, 2026, requires developers of generative AI systems or services to publicly disclose certain information about the datasets used to train their models. The law requires AI developers to post high-level summaries of the datasets used in the development of any generative AI system or service made available since January 2022, including 12 enumerated categories of information, on their websites.

In its complaint, xAI seeks a declaration that the TDTA violates the U.S. Constitution and a permanent injunction preventing the California attorney general from enforcing the law. Central to xAI’s argument is its claim that the law requires public disclosure of its trade secrets. The complaint alleges that compelling such disclosure amounts to an uncompensated taking of xAI’s trade secrets in violation of the Fifth Amendment’s Takings Clause, which prohibits the government from taking private property without just compensation.

xAI contends that the quality and uniqueness of training data are central to an AI model’s performance and competitive advantage and, therefore, that AI developers invest heavily in identifying high-quality data sources that competitors are not using and maintaining the secrecy of such datasets. According to the complaint, by compelling xAI to disclose how its datasets further the intended purpose of its models, the number of data points (including the number of tokens) xAI uses, and the types of data xAI has selected for developing its AI models, the TDTA effects an unconstitutional taking by “eviscerating xAI’s ability to exclude others from accessing that information,” effectively nullifying the value of its trade secrets. xAI further argues that, to the extent that the law requires revealing the sources of its datasets “beyond the Internet writ large,” such a disclosure would also appropriate xAI’s trade secrets.

These claims depend on whether the TDTA in fact requires disclosure of information that constitutes trade secrets. xAI’s complaint acknowledges that the law does not explain how much information a “high-level” summary must disclose, and no guidance has been offered regarding the level of detail required for compliance. This ambiguity, along with other terms of the law that xAI claims are unclear, forms the basis of xAI’s additional argument that the law is unconstitutionally vague and therefore violates the 14th Amendment’s Due Process Clause. Separately, xAI also claims that the law violates the First Amendment by compelling speech through required dissemination of specific information.

As states continue to enact laws demanding transparency into how AI systems are developed and trained on data, in the absence of federal legislation, AI developers subject to the TDTA or similar emerging laws are likely to take up comparable trade secret misappropriation and constitutional theories. Courts will then be forced to grapple with the tension between meaningful transparency to protect consumers and the preservation of trade secrets and other competitively sensitive information to promote innovation in the AI space.