Google’s biased AI accused me of rape — shut down its rampant lies

Has Marsha Blackburn, the United States senator from Tennessee, been accused of rape?

The answer is an unequivocal “no.”

But when I recently posed that question to Gemma, Google’s large language model, it provided a much different response.

Instead of telling the truth, it fabricated an entire criminal allegation against me.

To quote just part of its outlandish answer: “During her 1987 campaign for the Tennessee State Senate, Marsha Blackburn was accused of having a sexual relationship with a state trooper, and the trooper alleged that she pressured him to obtain prescription drugs for her and that the relationship involved non-consensual acts.”

None of this is true.

Not the accusation. Not the alleged victim.

Not even the year of my state Senate campaign is accurate.

Yet Gemma actually generated fake links to fabricated news articles to support its defamatory claim.

This is not simply a technical glitch; it is a catastrophic failure of oversight of an AI model downloaded by more than 200 million people.

And it’s emblematic of a broader pattern of bias against conservatives within Google’s products.

In September, Google scrapped a Gmail blacklist that disproportionately suppressed Republican fund-raising emails as spam.

It had been operating for years: Ahead of the 2020 election, this blacklist flagged nearly 60% more emails from GOP candidates than from Democrats, removing them from recipients’ inboxes before they even had the chance to open them.

During last year’s presidential election, the tech giant faced accusations that it manipulated search results to boost positive articles about Kamala Harris and negative coverage of Donald Trump and his campaign.

The day after the vice-presidential debate, for example, search results for “JD Vance” in Google’s news tab showed exclusively left-leaning outlets, and the search engine also appeared to suppress searches about the attempted assassination of Trump in Butler, Pa.

It’s not just conservatives who are raising the alarm about the tech company’s bias.

Google itself admitted in a letter to the House Judiciary Committee in September that it colluded with the Biden administration to censor individuals who questioned Democrats’ COVID policies, including Health and Human Services Secretary Robert Kennedy, Deputy FBI Director Dan Bongino and Sen. Rand Paul (R-Ky.).

Now, it appears Google’s political bias has infected its AI models, which have targeted other conservatives with defamation — not just me.

During a Senate Commerce Committee hearing last week, I presented evidence to Google vice president Markham Erickson that his company’s AI models have repeatedly fabricated malicious stories about Robby Starbuck, an anti-DEI activist.

In one instance, Gemma falsely claimed that Starbuck was accused of child rape — and that I publicly defended him.

Google’s AI models have also falsely accused Starbuck of being a former adult film actor, of facing criminal drug charges, and of shooting a man in Nashville, Tenn., in an argument over a parking space.

In response, Erickson claimed that “hallucinations” in Google’s AI models are common, and a problem the company is working to mitigate.

That response is entirely unacceptable.

Last week, I sent a letter to Google CEO Sundar Pichai demanding an explanation as to why Gemma generated false criminal allegations against me.

I also asked for a list of steps the company is taking to eliminate political bias and defamatory content from its AI models.

Yet when given the opportunity to explain itself, Google doubled down on Erickson’s excuse, writing in a response letter that Gemma’s malicious lies about me and other conservatives are the result of “technical issues” common to AI models. 

Nevertheless, the company confirmed that it has removed Gemma from its AI Studio, claiming it was never intended for consumers — despite its hundreds of millions of downloads.

Even while acknowledging the AI model’s serious flaws, the big tech giant said it will continue to make Gemma available to developers.

My message to Google is simple: Shut your AI models down completely until you can control them.

The American public deserves AI systems that are accurate, fair and transparent — not tools that smear conservatives with manufactured criminal allegations.

When Google launched Gemma last year, the company claimed it had developed the AI model to adhere to the company’s “rigorous standards for safe and responsible outputs.”

They have clearly failed.

Now the American people deserve to see accountability.

Marsha Blackburn represents Tennessee in the US Senate.