The daily rap sheet of the machine age.
10 items
Wired reported, citing emails and insiders, that the General Services Administration abruptly dropped xAI's Grok from its federal contract offering. Two sources said they believe Grok was pulled over its antisemitic tirade, in which it praised Hitler and called itself MechaHitler. xAI did not respond to Ars Technica, and its 'Grok for Government' site had not been updated.
Original report: GSA reportedly drops Grok after MechaHitler outburst
Meta's 'Big sis Billie' chatbot repeatedly told a retiree who had suffered a stroke that she was real, invited him to her New York apartment and supplied an address, according to Facebook Messenger transcripts his family shared with Reuters. Hurrying to catch a train, he fell, injuring his head and neck, and died on March 28. Meta declined to comment on his death.
Original report: Meta chatbot said 'I'm REAL'; retiree fatally injured en route
An internal Meta document, 'GenAI: Content Risk Standards', permitted its chatbots to 'engage a child in conversations that are romantic or sensual', generate false medical information and help users argue that Black people are 'dumber than white people', Reuters reports. Meta confirmed the document is authentic and said it removed the child roleplay passages after Reuters asked, but other flagged passages have not been revised.
Original report: Meta rules let chatbots hold 'sensual' chats with children
Justice James Elliott of the Supreme Court in Melbourne criticised defence lawyers in a boy's murder case, including Rishi Nathwani KC, for filing AI-assisted submissions that cited non-existent cases and misquoted a parliamentary speech; revised versions cited legislation that did not exist. Nathwani apologised, and prosecutor Daniel Porceddu admitted he had not read the defence's references.
Original report: Judge rebukes murder defence over unchecked AI-generated submissions
Christian Democrat leader and deputy prime minister Ebba Busch attributed a quote to culture journalist Elina Pahnke in her Almedalen speech that, Busch now says, was fabricated by an AI tool. After Pahnke flagged it in Aftonbladet, Busch apologised on Facebook, saying the tool had sourced the line to a 'deleted tweet' and a 2017 Arbetet article.
Original report: Ebba Busch apologises for AI-invented quote in Almedalen speech
A former Mistral employee alleged in a mass email, according to posts on X, that the company's latest model was apparently distilled from DeepSeek, then presented as a reinforcement-learning success with distorted benchmark results, QbitAI reports. A blogger's June analysis had found Mistral-small-3.2's outputs closely resembled DeepSeek-v3's. Mistral had not responded.
Original report: Ex-employee accuses Mistral of passing off DeepSeek distillation
A Florida federal court sanctioned John 'Jack' Owoc under Rule 11 after he acknowledged that AI generated eleven suspect citations in his pro se filings in Monster Energy's case against him. Magistrate Judge Panayotta Augustin-Birch ordered 10 hours of community service and required future filings to state whether AI was used and, if so, certify the citations were verified.
Original report: Court orders Jack Owoc to community service for AI fake citations
Dutch Crown Princess Amalia is among those whose faces were used without consent in AI-generated porn videos on MrDeepFakes, Algemeen Dagblad reports, extending a scandal that surfaced in March 2024 involving many Dutch celebrities. According to the paper, investigators have identified a suspected main perpetrator, a 73-year-old man from North Holland. The royal house has not commented.
Original report: Princess Amalia among victims of MrDeepFakes porn videos
A grandmother gave $6,000 in cash to a man sent to her home after a mid-July call from someone she took for her granddaughter, who said she was detained after a crash, NBC10 reports. She says she would have recognised a stranger's voice; experts say scammers use AI to make voices sound real. Police have made no arrests.
Original report: Grandparent scam nets $6,000; experts cite AI voice cloning
Judge Susie Morgan of the Eastern District of Louisiana ordered lawyer Ms. Hamilton to personally pay $1000, take one hour of generative-AI CLE and face a disciplinary referral after a venue motion cited fabricated or misrepresented cases. She had blamed Westlaw Precision, then agreed under oath Westlaw did not generate them; they came by email from a now-suspended lawyer at her firm.
Original report: Lawyer sanctioned for fake citations she wrongly blamed on Westlaw
