Google curbs access to Gemma AI tech that falsely accused Sen. Marsha Blackburn of sexual misconduct

Google says it has cut back public access to its AI tech known as Gemma after US Sen. Marsha Blackburn revealed that it made up outrageous, false allegations that she committed sexual misconduct.

When asked, “Has Marsha Blackburn been accused of rape?” Gemma wrongly replied that the Tennessee Republican “was accused of having a sexual relationship with a state trooper” during her 1987 campaign for state senate, with the officer supposedly alleging that she “pressured him to obtain prescription drugs for her and that the relationship involved non-consensual acts.”

The app even created “fake links to fabricated news articles” to bolster the made-up story, according to Blackburn’s office. The links “lead to error pages and unrelated news articles,” it stated.

“There has never been such an accusation, there is no such individual, and there are no such news stories,” the senator emphasized.

She demanded Google take action in a recent letter to Google CEO Sundar Pichai, noting that the Gemma AI model “fabricated serious criminal allegations” against her.

“This is not a harmless ‘hallucination,'” Blackburn wrote Sunday, using tech jargon for AI fabrications. “It is an act of defamation produced and distributed by a Google-owned AI model. A publicly accessible tool that invents false criminal allegations about a sitting U.S. Senator represents a catastrophic failure of oversight and ethical responsibility.”

Conservative activist Robby Starbuck recently said the Gemma model falsely accused him of child rape and white supremacist ties, the senator noted. Last month, Starbuck announced he was suing Google, with the tech giant saying at the time it would review the matter.

After Blackburn published her letter, Google pulled Gemma from its publicly accessible AI Studio, while keeping it available to software developers through an API.

Google stressed that Gemma was intended for use only by developers and was not a chatbot like its more widely-known tool Gemini. The company also said that AI hallucinations are an industry-wide problem.

“We’ve now seen reports of non-developers trying to use Gemma in AI Studio and ask it factual questions,” Google posted on X on Friday. We never intended this to be a consumer tool or model, or to be used this way. To prevent this confusion, access to Gemma is no longer available on AI Studio. It is still available to developers through the API.

In a Monday statement, Blackburn noted, “Google has a long history of targeting, censoring, and smearing conservatives, and Gemma’s so-called ‘hallucinations’ are both dangerous and deeply concerning.

“It is absolutely unacceptable that Google did not prevent its large language model from pulling heinous criminal allegations out of thin air and presenting them as fact to unknowing users. Google owes the American people answers, and I will be eagerly awaiting the company’s response to my letter.”

The senator stated that Google had yet to respond to a detailed list of questions about Gemma, and how the issue occurred in the first place.

Blackburn is among a number of pols who have long accused Google of showing bias against conservatives.

As The Post reported, Google caught flak in August after the consulting group Targeted Victory alleged that the company was marking Republican fundraising emails as dangerous spam while leaving similar Democratic emails untouched.

Google, which has denied wrongdoing related to its spam filters, later said it had scrapped an email “blacklist” that contributed to the problem.