AI Accuses Journalist of Escaping Psych Ward, Abusing Children and Widows

Hello!

Microsoft’s AI chatbot Copilot falsely accused a journalist of crimes he had covered in his reporting, The Register reports. The incident adds to the growing list of cases where AI hallucinations generate damaging and entirely false information.

The journalist in question, Martin Bernklau, a court reporter based in Tübingen, Germany, discovered that the AI had portrayed him as an escapee from a psychiatric ward, a convicted child molester, and a fraudster who targets grieving widows.

German public broadcaster Südwestrundfunk, which first reported the story, noted that Copilot also published Bernklau’s full name, address, phone number, and even a route planner to his home.

“This seriously violates my human dignity,” Bernklau told SWR.

No Recourse

Bernklau stumbled upon the false claims while checking how his own court reports appeared online. He queried a version of Copilot integrated into Microsoft’s Bing search engine and received the fabricated accusations.

The chatbot had apparently merged details from Bernklau’s decades of coverage of criminal trials and incorrectly presented him as the perpetrator.

Bernklau filed a criminal complaint for defamation, but prosecutors declined to pursue the case. They concluded that no individual could be held responsible for the AI-generated statements, according to SWR.

He achieved limited success when he contacted data-protection authorities. Microsoft assured Bavaria’s data-protection officer that the fabricated content would be removed. However, the material reappeared after only three days. For three months, Bernklau observed that his name was sometimes blocked and at other times accessible, with the situation changing daily or even hourly.

Prolific Liar

This is far from an isolated incident. In 2025, Meta’s AI chatbot falsely accused a Stanford AI researcher of terrorism. More recently, Elon Musk’s Grok chatbot wrongly implicated an NBA player in a series of vandalism attacks, likely due to misread joke posts on social media.

These examples reflect a wider pattern of AI systems producing misinformation—and sometimes disinformation—that can inflict serious personal harm when individuals are specifically targeted.

Bernklau described the experience as deeply traumatic, a “mixture of shock, horror, and disbelieving laughter.” He called it “too crazy, too unbelievable, but also too threatening.”

Whether Microsoft can be held legally responsible for statements made by its chatbot remains unresolved. Ongoing lawsuits, including a case brought against OpenAI after ChatGPT made similar false accusations of embezzlement, may eventually establish legal precedent. For now, Bernklau has few practical remedies.

Thank you!
Join us on social media!
See you!