Google pulls Gemma AI model after US senator says it fabricated assault claim

Google has removed its open‑weight Gemma model from AI Studio after US Senator Marsha Blackburn said the system invented a sexual assault allegation about her, prompting a formal complaint to CEO Sundar Pichai.

Google kept Gemma available to developers via API while it curtailed access in the public Studio interface.

In statements and posts, Google said non‑developers have been using Gemma in AI Studio for factual Q&A, which the company did not intend. To “prevent confusion,” access in AI Studio has been removed, though developers can still call Gemma through the API.

Blackburn alleges defamation

Senator Blackburn has written to Pichai alleging that Gemma fabricated serious misconduct claims and supported them with non‑existent links, and she pressed Google for an explanation.

Her letter and public posts arrived shortly before and after Google changed Gemma’s availability.

Google said hallucinations are an industry‑wide challenge, especially for smaller open models, and that AI Studio was never meant as a consumer venue for fact‑checking. It therefore pulled Gemma from AI Studio to reduce misuse while continuing to improve safeguards.

The move has come amid heightened scrutiny of generative AI accuracy. In a recent, separate case, conservative activist Robby Starbuck sued Google alleging that AI systems produced defamatory falsehoods; the company acknowledged hallucinations as a known risk while saying it is working to minimise them.