Google has taken the significant step of removing its AI model, Gemma, from the AI Studio platform following an accusation from U.S. Senator Marsha Blackburn. The senator contended that Gemma produced false and defamatory statements regarding her when prompted with a specific question about her past.
In her correspondence to Google CEO Sundar Pichai, Blackburn highlighted that Gemma generated a fabricated narrative claiming that during a 1987 state senate campaign, she was allegedly involved in serious misconduct. This included assertions that a state trooper accused her of pressuring him for prescription drugs and of being involved in non-consensual acts. Blackburn firmly refuted these claims, stating that not only were they untrue, but also that the timeline was incorrect as her campaign occurred in 1998. She emphasized there have been no such allegations against her and pointed out that the references provided by Gemma led to broken links or irrelevant articles.
Blackburn characterized the model’s output as “an act of defamation” rather than a mere “harmless hallucination,” which she argued has potentially damaging implications for individuals when AI-generated content spreads misinformation.
This incident falls closely on the heels of a recent Senate Commerce hearing, where Blackburn addressed the issue of AI-generated misinformation directly with Google’s Vice President of Government Affairs, Markham Erickson. During the hearing, Blackburn brought attention to a lawsuit filed by conservative activist Robby Starbuck, which alleges that Google’s AI models, including Gemma, disseminated false claims, labeling him as a “child rapist” and “serial sexual abuser.” Erickson acknowledged the challenges associated with AI hallucinations but stated that efforts are underway to combat these inaccuracies.
In the wake of these criticisms, Blackburn pointed out concerns raised by supporters of former President Donald Trump regarding a perceived bias against conservative figures manifesting within Google’s AI systems.
In a statement posted on X (formerly Twitter), Google did not directly address Blackburn’s letter but mentioned they had observed non-developers using Gemma in ways they did not intend, stressing that it was designed solely as a developer tool rather than for consumer use. Consequently, Google has decided to retract Gemma from the AI Studio web platform, though it will still be accessible through its API for developers.
This situation underscores the ongoing challenges faced by technology companies as they navigate the ethical implications of AI development, particularly regarding misinformation and bias in its applications. The proactive steps taken by Google in response to these concerns could serve as a pivotal moment in ensuring the responsible use of AI.