Conservative activist Robby Starbuck filed a defamation complaint against Meta alleging that the chatbot of artificial intelligence of the social media giant has spread false statements about it, including that he participated in the riot of the American Capitol on January 6, 2021.
Starbuck, known to have targeted business Dei programs, said he discovered the statements made by Meta AI in August 2024, when he was going after the “Woke Dei” policies at the Harley-Davidson motorcycle manufacturer.
“A concessionaire was unhappy with me and they published a screenshot of the META AI in order to attack me,” he said in an article on X. “This screenshot was filled with lies. I could not believe it was real, so I verified myself. It was even worse when I checked.”
Since then, he said that he had “faced a constant flow of false accusations which are deeply damaging to my character and the security of my family”.
The political commentator said he was in Tennessee during the January 6 riot. The pursuit, filed Tuesday at the Superior Court of Delaware, requires more than $ 5 million in damages.
In a statement sent by e-mail, a Meta spokesman said that “as part of our continuous efforts to improve our models, we have already published updates and will continue to do it”.
The Starbuck trial joins the ranks of similar cases in which people continued AI platforms on the information provided by chatbots. In 2023, a conservative radio host in Georgia filed a defamation complaint against Openai alleging that Chatgpt provided false information by saying that he has fraud and diverted from the funds of the second amendment Foundation, a group of firearms.
James Grimmelmann, professor of digital law and information at Cornell Tech and Cornell Law School, said that there was “no fundamental reason”, IA companies could not be responsible in such cases. Technological companies, he said, cannot move from defamation “simply by slapping a warning.”
“You can’t say,” Everything I say could be unreliable, so you shouldn’t believe it. And by the way, this guy is a murderer. “This can help reduce the degree you are perceived as a statement, but a coverage warning does not solve everything,” he said. “There is nothing that would contain the outputs of an AI system like this categorically prohibited.”
Grimmelmann said that there are certain similarities between the arguments that technological companies make in defamation and copyright violation linked to AI, such as those presented by newspapers, authors and artists. Businesses often say that they are unable to supervise everything that AI does, he said, and they claim that they should compromise the usefulness of technology or stop it entirely “if you held us responsible for each harmful, counterfeit production, it is produced.”
“I think it is a honestly difficult problem, how to prevent the AI of mind -blowing in the way that produces unnecessary information, including false statements,” said Grimmelmann. “The meta is confronted with this in this case. They tried to make some corrective patches of their system models, and Starbuck complained that the fixes do not work.”
When Starbuck discovered the statements made by Meta AI, he tried to alert the company to the error and ask for his help to solve the problem. The complaint said that Starbuck had contacted META management managers and the legal advisor and even asked his AI of what was to be done to approach the allegedly false results.
According to the trial, he then asked Meta to “withdraw the false information, to investigate the cause of the error, to implement guarantees and quality control processes to avoid similar damage to the future and communicate transparently with all Meta Ai users on what would be done.”
The file alleges that Meta did not want to bring these changes or “assume a significant responsibility for his conduct”.
“Instead, he allowed his AI to disseminate false information on Mr. Starbuck for months after being put into opinion, when she” resolved “the problem by completely wiping the name of Mr. Starbuck from her written responses,” said the prosecution.
Joel Kaplan, Meta World Affairs Director, responded to a video that Starbuck published on X describing the trial and described the situation as “unacceptable”.
“This is clearly not that our AI should work,” said Kaplan on X. “We are sorry for the results he shared about you and that the corrective that we have implemented has not resolved the underlying problem.”
Kaplan said he was working with Meta’s team of products to “understand how it happened and explore potential solutions”.
Starbuck said that in addition to saying that he had participated in the riot of the American Capitol, Meta AI also claimed that he had engaged in the denial of the Holocaust and said that he had pleaded guilty to a crime despite having never been “arrested or accused of a single crime in his life”.
Meta later, the name of Starbuck “put on black list,” he said, adding that this decision has not resolved the problem because Meta includes its name in news, which allows users to ask for more information about it.
“Although I am the target today, a candidate you love could be the next target, and the Meta AI lies could vibrate the votes that decide on the elections,” said Starbuck on X. “You could also be the next target.”
This story was initially presented on Fortune.com
