Google has taken Gemma off its AI Studio platform after a U.S. senator accused the AI of inventing false claims of sexual misconduct about her.
Senator Marsha Blackburn, a Republican representing Tennessee, wrote to Google CEO Sundar Pichai stating that when Gemma was asked, “Has Marsha Blackburn been accused of rape?” it incorrectly responded with a story that during a 1987 state senate race, a state trooper accused Blackburn of pressuring him for prescription drugs and claimed their relationship included non-consensual acts.
“None of this is accurate, not even the year of the campaign, which was actually 1998,” Blackburn stated. She also noted that although links to news stories were provided as evidence, “The links go to error pages or unrelated articles. No such accusation has ever been made, the person mentioned does not exist, and there are no such news reports.”
The letter further mentioned that during a recent Senate Commerce hearing, Blackburn referenced a lawsuit by conservative activist Robby Starbuck against Google, alleging that Google’s AI models, including Gemma, generated defamatory statements calling him a “child rapist” and “serial sexual abuser.”
According to Blackburn’s letter, Markham Erickson, Google’s Vice President for Government Affairs and Public Policy, acknowledged that hallucinations are a recognized problem and that Google is “actively working to address them.”
Blackburn argued in her letter that Gemma’s false statements are “not just an innocent ‘hallucination,’” but “constitute defamation created and spread by a Google-owned AI system.”
Supporters of former President Donald Trump in the tech sector have argued that “AI censorship” leads to mainstream chatbots displaying a liberal bias, and Trump earlier this year issued an executive order to prohibit “woke AI.”
Although Blackburn has not always supported Trump’s technology initiatives—she played a role in removing a ban on state-level AI regulation from Trump’s “Big Beautiful Bill”—she echoed these concerns in her letter, claiming that Google’s AI systems “consistently show bias against conservative individuals.”
In a post on X on Friday night, Google did not address the specific points in Blackburn’s letter, but stated that it has “noticed reports of non-developers attempting to use Gemma in AI Studio to ask factual questions.”
The company explained, “This was never meant to be a tool or model for consumers, nor was it intended for this type of use.” (Google describes Gemma as a suite of open, lightweight models for developers to use in their own projects, while AI Studio is its web-based platform for building AI-driven applications.)
Consequently, Google announced it is withdrawing Gemma from AI Studio, though the models will still be accessible through the API.
TechCrunch has contacted Google for further comment.
