A Delaware judge has refused to throw out a defamation suit over false statements generated by Google's AI chatbots, allowing the first such case to proceed to discovery.
The ruling
Judge Meghan A. Adams of the Delaware Superior Court denied Google's motion to dismiss in Starbuck v. Google LLC, case number N25C-10-211 MAA. She found that publication, actual malice and emotional-distress damages had all been adequately pled, and described the dispute as a new frontier for defamation law.
What the complaint alleges
Conservative activist Robby Starbuck filed in October 2025, identifying 13 categories of false statements produced by Bard and later Gemini — among them fabricated sexual assault accusations, a false association with white supremacist Richard Spencer, stolen valor and involvement in the Capitol riot. He seeks at least $15 million.
What this is not
Google has not lost. A denied motion to dismiss means the allegations are sufficient to proceed, not that any of them are proven. The $15 million is Starbuck's demand, not an award. And this is Delaware's Superior Court — a state trial court, not a federal court and not the Court of Chancery.
The doctrine problem
Defamation requires a false statement of fact, published, with the required degree of fault. Each element strains against a system that answers differently every time it is asked. Who published — the model, the company, or the user who prompted it? And actual malice, which demands knowledge of falsity or reckless disregard for it, has never been applied to a probabilistic text generator. Judge Adams did not resolve those questions; she held that they cannot be resolved on a motion to dismiss.
What discovery could surface
The reason this matters more than the pleading standard: discovery can compel Google to produce internal material on what it knew about its models' propensity to fabricate, and when.