Google has temporarily removed its speech model Gemma from its AI Studio environment after US Senator Marsha Blackburn accused the AI of spreading false claims, TechCrunch reports.
In a letter to Google CEO Sundar Pichai, the Tennessee senator said that during the test, Gemma provided false information, saying that she had been accused of rape in 1987. The model even provided fabricated links to news stories that led to nonexistent pages.
Blackburn stressed that such events never happened, and called the model's response "an act of defamation created and spread by Google's AI." She also mentioned a lawsuit by activist Robbie Starbuck, who accuses Google of spreading similar fabrications about him.
During a Senate hearing, Google's vice president of public policy, Markham Erickson, acknowledged that "hallucinations" are a known problem with AI models, and the company is working to minimize it.
After the letter was published, Google said that Gemma was not intended for consumer use and was created as a tool for developers. The company temporarily removed the model from AI Studio, but left it available through the API.