Senate Republicans Criticize Google’s AI for Allegedly Defaming Conservatives

Senate Republicans have accused Google and its artificial intelligence of unfairly targeting conservatives with false claims and fake news. This includes serious allegations like fabricated sexual assault accusations that aren’t based in reality.

Senator Marsha Blackburn from Tennessee addressed a letter to Google’s CEO, Sundar Pichai, bringing attention to Google’s language model AI, Gemma, which she suggested made defamatory claims against her, including a false sexual assault allegation.

Blackburn emphasized that this AI generated a fictitious story about her, linking her to an unsupported sexual assault claim, alongside misleading articles that purported to back up this falsehood.

Her correspondence with Pichai followed a recent Senate Commerce Committee hearing that explored “jawboning.” This practice involves government figures exerting indirect pressure on tech companies like Google to censor information on platforms.

During the hearing, Blackburn challenged Markham Erickson, Google’s VP of government affairs, over an instance where the AI supposedly created false accusations targeting conservative activist Robbie Starbuck.

“Hallucinations” from AI refer to when generative models, such as Gemma, produce misleading or inaccurate data presented as facts.

Starbuck, who is currently suing Google, claims that the AI was responsible for linking him with grave accusations, including sexual assault and child exploitation.

This led Blackburn to query the AI directly, asking if there were allegations against her. The response was a fabricated article insinuating she had a sexual relationship with a state trooper in 1987 while campaigning for the Tennessee Senate, suggesting coercive behavior involving medication. Blackburn pointed out that she actually started her political career in 1998 and stated, “There’s never been an accusation like that, there’s never been a person like that, there’s never been a news story like that.”

She argued that such “hallucinations” are not trivial but constitute defamation. Blackburn emphasized the gravity of this situation, describing it as a significant failure of oversight and ethical responsibility from Google.

Moreover, she claimed there is a recurring pattern of bias within Google’s AI against conservatives. Whether intentional or simply resulting from biased training data, she argued, the outcome remains detrimental, as it amplifies false narratives that undermine public trust.

Blackburn has requested that Google clarify by November 6 how Gemma produced these defamatory claims, the measures in place to mitigate political bias in its AI, what safeguards failed, and how the company plans to resolve the defamatory content and prevent incidents like this from happening again.

During the hearing, Erickson referenced large language models’ tendency to “hallucinate.” Blackburn expressed her discontent, insisting on strict measures until the issues are addressed. As of now, Google has yet to respond to inquiries regarding this matter.