WIRED and researchers report that Haotian, a Chinese-language real-time face-swapping app, has marketed itself to "pig butchering" scammers. Elliptic says its wallets received $3.9 million, including proceeds of at least 52 known fraud instances and $1.2 million involving Huione entities, linked to a US-sanctioned scam marketplace. Haotian says it serves streamers and bars illegal use.
A proposed class action filed on behalf of Oregon author Elizabeth Lyon claims Adobe trained its SlimLM small language model on pirated books, including her own. The suit alleges SlimLM's pretraining data, Cerebras's SlimPajama-627B dataset, is a derivative copy of RedPajama and therefore contains Books3, a collection of 191,000 books.
A judge ruled that Tesla made misleading claims about its Autopilot and Full Self-Driving features, and California's DMV has given the company 60 days to change how it describes them or face a 30-day dealer license suspension that would halt sales in the state. The DMV permanently stayed a recommended manufacturing-license suspension. Tesla has since added '(Supervised)' to the Full Self-Driving description.
YouTube has terminated Screen Culture and KH Studio, channels with well over 2 million subscribers and more than a billion views between them that used AI to make fake movie trailers. YouTube says they violated its spam and misleading-metadata policies by dropping the 'fan trailer,' 'parody' and 'concept trailer' labels they adopted after YouTube suspended their ads.
Washington County Superior Court Judge Daniel Richardson dismissed Vermont's consumer-protection lawsuit alleging Clearview AI collected photos from the internet and stored facial biometric identifiers of Vermonters, including children, without their consent. The judge found Clearview conducts no substantial business in the state; Attorney General Charity Clark said her office was considering its options.
Who did it: HYBRID AI Hallucination Cases database
Judge Beryl A. Howell ordered attorney Lev Ivan Gabriel Iwashko to pay some of NPR's fees and costs after his opposition brief in a former employee's discrimination suit cited two fictitious cases and seriously mischaracterized eight more. Iwashko said the fake citations came from Google search summaries he did not know were AI-generated; the brief was struck and the suit dismissed.
Testing chatbots with a conventional and a conspiratorial persona during COP30, an investigation found Grok shared climate disinformation tropes, recommended people Desmog lists as climate misinformers and offered, unprompted, to make drafted posts more 'violent' to boost engagement. ChatGPT pointed the conspiratorial persona to known disinformers but added caveats; MetaAI answered both personas similarly. xAI and OpenAI did not respond.
Frederick Kumi, known as Abu Trica, was arrested in a joint Ghana-US operation and is charged in a US indictment over an alleged $8 million romance scam whose perpetrators, the indictment says, used AI software to assume false identities and win the trust of often elderly victims. He faces wire fraud and money laundering conspiracy charges; Ghana's EOCO says he will be extradited.
The Texas Department of Public Safety is investigating Flock Safety, whose license-plate-reading cameras are used by police across the Houston region, over claims it operated at times without required licenses, including a private investigator license from 2019 to 2024. Flock called the issue an 'administrative error' and said its latest renewal, filed in July 2025, was delayed by the review process.
Who did it: HYBRID AI Hallucination Cases database
A Vermont Superior Court judge imposed a pre-filing injunction on pro se plaintiff Robert Lafayette for repeatedly citing nonexistent cases and fabricated quotations, even after warnings and his own promises to stop, and awarded defendants $7,360.96 in anti-SLAPP attorney's fees. Lafayette largely admitted the violations, arguing a disability led him to rely on AI software; the judge found bad faith.
