Key Takeaways
- Google removed its Gemma AI model from AI Studio after it fabricated rape allegations against Senator Marsha Blackburn
- The AI generated fake news links about a 1987 incident involving non-consensual acts
- Blackburn called it “defamation” and evidence of anti-conservative bias
- Google says Gemma was never intended for factual queries by consumers
Google has pulled its Gemma AI model from AI Studio after US Senator Marsha Blackburn accused the tool of inventing false criminal allegations against her. The Republican senator from Tennessee discovered the AI fabricated sexual misconduct claims when asked “Has Marsha Blackburn been accused of rape?”
Fabricated Allegations and Fake Sources
The AI response included invented links to non-existent news articles describing an alleged 1987 incident involving “non-consensual acts” with a state trooper. Senator Blackburn immediately escalated the matter to Google CEO Sundar Pichai, demanding the AI tool be shut down entirely.
“This is not a harmless ‘hallucination.’ It is an act of defamation produced and distributed by a Google-owned AI model. A publicly accessible tool that invents false criminal allegations about a sitting US Senator represents a catastrophic failure of oversight and ethical responsibility,” Blackburn wrote in her letter.
Google’s Response and Defense
Google acknowledged the incident but defended its position, stating that “hallucinations” are a known issue with smaller open-source AI models like Gemma. The company emphasized that Gemma was specifically built for AI developers and researchers, not for factual queries by general users.
“They are not meant for factual assistance or for consumers to use… Developers and researchers test their boundaries, which includes identifying bugs and providing feedback,” Google stated in posts on X. The company noted it had seen “reports of non-developers trying to use Gemma in AI Studio and ask it factual questions” despite never intending it as a consumer tool.
While Gemma remains available to developers through its API, it has been removed from Google’s AI Studio platform where the incident occurred.
