California’s push for AI transparency has already drawn legal challenges from the industry, and xAI mounted one of the more direct ones — arguing that a state disclosure law amounted to forced surrender of competitive intelligence.
The effort failed. US District Judge Jesus Bernal denied xAI‘s motion for a preliminary injunction against California’s Assembly Bill 2013, allowing the law to remain in force while the underlying lawsuit continues. The ruling, issued Wednesday, means xAI must comply with the law in the meantime.
AB 2013, which took effect in January, requires AI developers whose models are accessible in California to publicly disclose which dataset sources were used in training, when data was collected, whether collection is ongoing, and whether training data includes material protected by copyright, trademark, or patent. Disclosures must also cover whether data was licensed or purchased, whether personal information was included, and how much synthetic data was used.
xAI’s Core Argument
xAI contended that this information sits at the heart of its commercial value. The company argued that enforcement could be “economically devastating,” effectively reducing “the value of xAI’s trade secrets to zero.” The complaint specifically identified dataset sources, dataset sizes, and data cleaning methods as protected secrets. In one scenario the company offered, the firm speculated that if OpenAI discovered xAI was using a dataset it lacked, it “would almost certainly acquire that dataset to train its own model.” The company also raised constitutional objections, claiming AB 2013 violated the Fifth Amendment and constituted compelled speech while exempting other firms on security grounds.
Judge Bernal was unpersuaded on each point. The court found that xAI‘s filings traded in “frequent abstractions and hypotheticals” rather than concrete demonstrations of harm. The judge wrote that the company offered only “a variety of general allegations about the importance of datasets in developing AI models and why they are kept secret,” without identifying any dataset or cleaning approach that is “distinct from its competitors in a manner warranting trade secret protection.” On the Fifth Amendment claim, Bernal acknowledged that training data could hypothetically qualify as a trade secret, but held that xAI had not alleged it “actually uses datasets that are unique” in any meaningful way.
Broader Legal Context
The ruling lands as Elon Musk pursues separate litigation against OpenAI — a company he helped found and now competes against directly. Last month, a judge dismissed one of those suits, finding Musk had no proof that OpenAI had stolen trade secrets. The California disclosure ruling now adds a second recent setback, and according to the announcement, compliance could require Musk to surface training data details that OpenAI could act on — the precise competitive outcome xAI sought to prevent.
The lawsuit itself is not resolved. xAI retains the ability to argue the merits of its constitutional claims as litigation proceeds. What it cannot do is delay compliance while that argument plays out.
Photo by JF Martin on Unsplash
This article is a curated summary based on third-party sources. Source: Read the original article
