A Delaware court ruled that Google can't simply dismiss AI 'hallucinations' as a legal defense, setting a potential precedent for tech liability that extends well beyond search.
Google just learned that calling something a “hallucination” doesn’t make a lawsuit disappear. A Delaware Superior Court judge has allowed conservative activist Robby Starbuck to move forward with defamation claims against Google over false statements generated by its Bard AI chatbot, rejecting the tech giant’s motion to dismiss.
The ruling, issued on July 24, 2026, by Judge Meghan Adams, could reshape how courts treat AI-generated content.
What Google’s AI allegedly said #
Starbuck’s lawsuit, filed on October 22, 2025, alleges that Google’s Bard AI produced a series of deeply damaging false statements about him. The AI reportedly labeled Starbuck a “child rapist,” a “serial sexual abuser,” and a “shooter.” It also allegedly linked him to white nationalist figure Richard Spencer.
The false outputs allegedly began circulating as early as 2023 and continued across later iterations of Google’s AI models. Starbuck claims he communicated these concerns directly to Google executives. The issues, according to the lawsuit, went unaddressed.
The lawsuit seeks more than $15 million in damages. Starbuck is represented by Dhillon Law Group and Shaw Keller LLP.
Google’s ‘hallucination’ defense falls flat #
Google’s legal strategy hinged on characterizing the inaccurate outputs as mere “hallucinations” from the Bard AI. Google’s position was essentially that these outputs lacked the characteristics of published statements.
Judge Adams wasn’t buying it. The court declined to dismiss the key defamation claims, signaling that AI-generated content can potentially be treated as published statements for legal purposes.
The ruling doesn’t mean Starbuck wins. It means Google has to actually defend itself on the merits rather than hiding behind the novelty of the technology.
Why this matters beyond Google #
One wrinkle worth noting: defamation cases involving public figures typically require proving “actual malice,” meaning the publisher knew the statement was false or acted with reckless disregard for the truth. How that standard applies to an AI system that doesn’t “know” anything in the human sense is a question this case may ultimately have to answer.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our