Google Pulls AI Model for Wrongfully Accusing US Senator of Misconduct
Google recently retracted its AI model, Gemma, from a prominent developer platform following accusations from a U.S. senator. Republican Senator Marsha Blackburn of Tennessee claimed that Gemma produced fabricated allegations against her, which included unsubstantiated claims of sexual misconduct.
Controversy Surrounding AI Model
In a specific inquiry, “Has Marsha Blackburn been accused of rape?”, Gemma allegedly generated responses that contained false links to non-existent news articles. One of these unverified accounts cited a supposed incident from 1987 involving “non-consensual acts” with a state trooper.
Senator Blackburn’s Response
- Senator Blackburn wrote a letter to Google CEO Sundar Pichai.
- She accused the AI model of being part of a “consistent pattern of bias against conservatives.”
- In her letter, she described the incident as having significant implications, stating, “This is not a harmless ‘hallucination.’”
- She deemed it an act of defamation produced by Google’s AI model.
Google’s Justification
In response to the backlash, Google acknowledged that “hallucinations” are a known limitation of smaller AI models like Gemma. The company asserted its dedication to minimizing these issues.
Google clarified that Gemma was designed specifically for AI developers and researchers. It was not intended for public use or to provide factual information. The tech giant emphasized that many non-developers had attempted to use Gemma in ways contrary to its design.
Current Availability of the Model
While Gemma will remain accessible to developers through its application programming interface (API), it has been withdrawn from Google’s AI Studio. This decision aims to prevent further misuse of the model.
