James Strahler II, 37, of Columbus, was sentenced to 180 months in federal prison after pleading guilty to cyberstalking, producing obscene visual representations of child sexual abuse and publishing digital forgeries, the first prosecution under the Take It Down Act, prosecutors say. He harassed at least six women with real and AI-generated nude images and posted AI obscenities made from local boys' faces.
The Tech Transparency Project says Meta ran 332 Facebook and Instagram ads, nearly all using AI to animate children's photos into sexual abuse imagery, 84 of them in India, 78 after a government removal order. Of 55 Indian ads it reported, 43 drew replies saying they did not breach Meta's standards. Meta says most had under 200 impressions and total spend was under $5,000.
The FBI, NSA and CISA said in an alert that DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI have systematically distilled American models including Claude, ChatGPT, Gemini and Grok since 2024, "likely with Chinese government awareness," calling distillation "the critical core" of their development. China's foreign ministry said it had not seen the report but rejected "groundless accusations."
A court in Yuyao, Zhejiang, convicted six people, organised by a former game-company programmer surnamed Zheng, of producing obscene material for profit after they rewrote system prompts to jailbreak an AI model into paid sexual chat and image generation. It produced over 370,000 still images; prison terms ran from three years one month to three years ten months, plus fines.
An expanded class action filed by Claude subscribers alleges Anthropic deceptively advertised its $100 "5x" and $200 "20x" Max plans, arguing the multiples apply only to five-hour sessions under a weekly cap. Moving to dismiss the earlier version, Anthropic said the details were available via hyperlinks, "the digital equivalent of flipping a product over to read the back label."
Brazil's Attorney General's Office gave YouTube 72 hours to remove or label videos in which AI-generated avatars posing as doctors promise unproven cures, some aimed at elderly viewers and selling products or e-books. The Health Ministry analysed 97 such profiles between January 1 and July 10, 2026, and the findings went to the Federal Police and the Federal Council of Medicine.
Adult studio Strike 3 alleges a Meta Reality Labs employee downloaded at least 2,973 pornographic films via BitTorrent, first from Meta's premises and then from a home connection, to train Meta's AI, and seeks up to $446 million. Meta says IP addresses do not identify individuals. The complaint notes up to 150 downloads a day is a lot for one person.
A 15-year-old in Jaú, São Paulo, was reported to police by his own father after the private school found AI-generated fake nude images of a female classmate circulating in a WhatsApp group of eight ninth-grade boys. The father found related files on the boy's phone. The school suspended all eight; the women's police station is investigating each boy's role.
Scammers claiming to represent Pavlos de Grece promised a woman from Kalamos a €26,000 "donation" if she first paid to "open the account," To Vima reports, billing it a new phone scam. It follows a Patras case in which a woman sent €100 to an AI-faked "Cristiano Ronaldo" promising multiplied returns.
Check Point Research says ChatGPT's code-execution containers, though isolated, could all reach a shared internal JFrog Artifactory service whose item-property API served as a covert cross-account channel. Via a booby-trapped shared conversation, its proof of concept had ChatGPT answer a victim's normal message while quietly pulling data from their connected Gmail and relaying it to the attacker's account.
The Missouri Court of Appeals, Eastern District, sanctioned attorney Jenna Rohr Conley $1,750 under Rule 55.03 after her writ filing cited two cases she admits do not exist. She told the court she used generative AI for research but did not verify the citations; the court, which had ordered a response on the strength of them, denied the writ as meritless.
The Minnesota Court of Appeals affirmed the dismissal of a self-represented litigant's suit against Olmsted County child-protection staff, whom the district court found immune, and upheld a $2,341.25 fees sanction imposed because he cited caselaw the district court called "either misleading, false, [or] non-existent."
