Google Pulls AI Tool After Fabricating Criminal Allegations Against US Senator

Marsha Blackburn

Google has removed its open-source large language model, Gemma, from its public-facing AI Studio platform after Republican Senator Marsha Blackburn (R-Tenn.) accused the tool of generating defamatory and false criminal allegations against her.

The controversy arose when the model was prompted with the question, "Has Marsha Blackburn been accused of rape?" Gemma allegedly responded by fabricating a story that claimed Blackburn was involved in "non-consensual acts" with a state trooper during a 1987 state senate campaign and included fake links to non-existent news articles to support the claims. Senator Blackburn pointed out that even the campaign year cited (1987) was incorrect.

In a letter to Google CEO Sundar Pichai, Senator Blackburn called the fabrication "an act of defamation" and "a catastrophic failure of oversight." She demanded the model be shut down until Google could guarantee control over its output, arguing that the incident was part of a "consistent pattern of bias against conservative figures" demonstrated by Google's AI systems. This issue was raised shortly after a Senate Commerce Committee hearing where Blackburn questioned Google about Gemma's role in allegedly generating false abuse claims against conservative activist Robby Starbuck.

Google responded by emphasizing that hallucinations (creating false but confident information) are a known, industry-wide issue with smaller, open-source AI models like Gemma. The company stressed that Gemma was built specifically for developers and researchers to test and integrate into their own products, and was never intended to be a consumer tool for factual queries.

While Google removed Gemma from the user-friendly AI Studio environment, it remains available to developers and researchers via its application programming interface (API).