The National Republican Senatorial Committee posted a 30-second ad on X and YouTube showing an AI-generated Chuck Schumer grinning and repeating "Every day gets better for us", a line he gave Punchbowl News only in text. The NRSC says it carries an AI disclaimer; Berkeley professor Hany Farid called the small watermark "not a particularly clear or obvious disclosure".
An appeals chamber in Esquel, Argentina, annulled a June 4 criminal sentence by judge Carlos Rogelio Richeri after spotting the line "Here you have point IV re-edited, without citations and ready to copy and paste" from a generative AI assistant, presumably ChatGPT. It voided the trial too, ordered a retrial before another judge and referred Richeri to Chubut's Superior Court of Justice for investigation.
Colombia's Superintendence of Industry and Commerce ordered World Foundation and Tools for Humanity, formerly Worldcoin, to immediately and permanently close data operations in the country and delete Colombians' iris biometric databases. The regulator found the firms, which offered cryptocurrency for iris scans, failed to clearly explain the data's purposes, secure it, or obtain express, prior, informed authorization.
Scale AI has agreed to settle four lawsuits filed in San Francisco Superior Court by former California workers who alleged they were misclassified as contractors, underpaid and denied benefits such as overtime and sick pay, according to court documents. Terms weren't disclosed and a judge must approve; one worker suit over exposure to disturbing content continues.
OpenAI said Thursday it has "paused" Sora users' ability to generate videos of Martin Luther King Jr., following backlash over "disrespectful depictions", and thanked Dr. Bernice A. King for reaching out on behalf of King, Inc. The company said estates and authorized representatives can request that a likeness not be used in Sora cameos.
Deputies arrested Moesha Gardener of North Bethesda, Maryland, on Oct. 10 and charged her with making a false statement to law enforcement over an allegedly fake Oct. 8 text telling her husband a man had forced his way into her house, followed by a photo of a man on their couch. Police say she called it a prank and said AI helped create the photo.
An AI-generated video circulated online showing George Freeman, Conservative MP for Mid Norfolk, announcing he had "joined Reform" and saying the "Conservative party had lost its way". Freeman called it a deepfake, said he has no intention of joining Reform or any other party, and reported it to Facebook and the police.
Neil Gardoll, 59, of Lalor Park in Sydney's west, was charged with importing "tier 2 goods" and with producing and possessing child abuse material after Border Force seized a child-like sex doll from a shipment. Investigators allege devices taken in a September 8 search held a "significant amount of AI-generated child abuse material".
Who did it: HYBRID AI Hallucination Cases database
Nebraska District Judge James M. Masteller fined lawyer James E. Harris $2,000, struck his pending motions and briefs, and required AI certifications after his filings cited nonexistent cases such as "Hogan v. Norfleet, 191 Neb. 123". Harris said AI output exceeded his protocols; the judge found his conduct "tantamount to bad faith" and directed that the order be sent to the Counsel for Discipline.
Who did it: HYBRID AI Hallucination Cases database
Poland's National Appeals Chamber ruled on 17 October 2025 that EXDROG's price explanations in a Kraków regional roads maintenance tender cited individual tax interpretations that did not exist. EXDROG did not deny using AI tools, noting no rule forbids it. The Chamber, unmoved, found the information misleading and ordered the contracting authority to exclude EXDROG and reject its offer.
A healthcare worker reported that Reddit Answers, the platform's AI chatbot, surfaced a post urging people to quit prescribed medication for high-dose kratom, and that, asked about heroin for pain, some answers encouraged it over prescriptions. After 404 Media picked up the story, Reddit said an update now stops related answers on sensitive topics appearing under posts.
NewsGuard tests found OpenAI's Sora 2 produced realistic videos advancing provably false claims for 16 of 20 prompts, including all five claims originating with Russian influence operations. A free online tool removed Sora's watermark in about four minutes. OpenAI said its usage policies prohibit misleading others through impersonation, scams or fraud.
