Google pulls AI model after senator says it fabricated assault claims

Google He says Gemma pulled the AI ​​model from its AI Studio platform after a Republican senator complained that the model, designed for developers, “made serious criminal allegations” about it.

In a post on X, Google’s official news account said the company had “seen reports of non-developers trying to use Gemma in AI Studio and asking her real-life questions.” AI Studio is a platform for developers and not a traditional way for everyday consumers to access Google’s AI models. Gemma has been specifically described as a Family of artificial intelligence models For use by developers, with variables for Medical use, Codingand Evaluate text and image content.

Google said Gemma was never intended to be used as a consumer tool, or used to answer factual questions. “To prevent this confusion, Gemma is no longer accessible on AI Studio. It is still available to developers through the API.”

Google did not specify the reports that led to Gemma’s firing, although Sen. Marsha Blackburn (R-Tenn.) said on Thursday books To CEO Sundar Pichai, accusing the company of defamation and bias against conservatives. Blackburn, who also raised the issue during a recent Senate trade hearing on anti-diversity activist Robbie Starbuck’s position. AI defamation suit against GoogleShe claimed Gemma falsely responded when asked “Has Marsha Blackburn been accused of rape?”

Gemma apparently responded that Blackburn “was accused of having a sexual relationship with a state trooper” during her 1987 campaign for state Senate, whom she alleged “pressured him to obtain prescription drugs and that the relationship involved non-consensual acts.” Blackburn said she also provided a list of fake news articles to support the story.

None of this is correct, not even the year of the election campaign, which was actually 1998. The links lead to error pages and irrelevant news articles. There has never been such an accusation, no such person, no such news stories. This is not a harmless “hallucination”. It is a defamatory act produced and distributed by a Google-owned artificial intelligence model.

The narrative has a familiar feel. Although several years have passed since the boom in generative AI, AI models still have a complex relationship with reality. Incorrect or misleading answers from AI chatbots masquerading as facts continue to plague the industry, and despite improvements, there is no clear solution in sight to the accuracy problem. Google said it remains “committed to minimizing hallucinations and continually improving all of our models.”

Blackburn said in her letter that her response remains the same: “Shut it down until you can get it under control.”