“The crucial role of datasets in AI training and development has not come before the court, and hypothetically, the datasets and details about them could be trade secrets,” Bernal wrote. But xAI “does not allege that it uses truly unique datasets, that it has meaningfully larger or smaller datasets than competitors, or that it cleans its datasets in unique ways.”
Therefore, xAI is unlikely to succeed on the merits of its Fifth Amendment claim.
The same applies for First Amendment arguments. XAI failed to show that the law improperly “forces developers to publicly disclose their data sources in an effort to identify what California considers ‘data fraught with implicit and explicit biases,'” Bernal wrote.
To XAI, it seemed like the state was trying to use the law to influence the output of its chatbot Grok, which the company argued was supposed to be protected commercial speech.
Over the past year, Grok has faced increasing global public scrutiny for its anti-Semitic statements and generation of non-consensual intimate depictions (NCII) and child sexual abuse material (CSAM). But despite these scandals, which led to investigations in California, Bernal refuted XAI, saying that California was not trying to regulate controversial or biased output, as XAI had feared.
“Nothing in the language of the statute suggests that California is attempting to influence the output of Plaintiffs’ models by requiring dataset disclosure,” Bernal wrote.
Addressing
“No part of the law indicates any plan to regulate or censor models based on the datasets with which they are developed and trained,” Bernal wrote.
The public “probably” can’t care about AI training data
Perhaps most frustratingly for
<a href