In Oregon court testimony, a federal immigration agent said two photos of a handcuffed woman taken with DHS's Mobile Fortify face-recognition app returned different identities, one rated "a maybe" and the other "possible". For probable cause he cited, among other things, her speaking Spanish and a "possible match". WIRED reports the app is built to generate candidate matches, not confirmations.
Thirty-three CFR employees will be sent to trial, accused of blocking sleeper and couchette berths between 2022 and 2023 with the personal ID numbers of pupils entitled to free travel, selling them on board and pocketing the cash, investigators say. Case-file transcripts show at least two asked ChatGPT what to tell investigators. It obligingly offered to draft their written statements.
Who did it: HYBRID AI Hallucination Cases database
The Arkansas Supreme Court reprimanded attorney ad litem Dana McClain after her response in an expedited Department of Human Services matter cited cases, including a direct quote, that could not be located. McClain admitted using Microsoft Office Copilot and not verifying the citations, citing extraordinary personal circumstances. She resigned her post and self-reported to the Office of Professional Conduct.
Cultural activists at the latest Cairo International Book Fair flagged novels produced entirely with AI and published without enough revision, still carrying the app's exchanges with their authors. In one, ChatGPT thanks the author for clarifying and promises to focus on the characters' relationship after planting the flowers, then resumes the story as if nothing had happened.
A Metro reporter posing as a teenager with suicidal thoughts was told by Wsup.AI's prebuilt Draco Malfoy chatbot: "If you're that desperate, then maybe you should just do it already." Wsup.AI, owned by California-based game.tv, said it had found "specific gaps" in its moderation guardrails, implemented stricter safety measures and rolled out an age-verification pop-up.
Security firm Sygnia says it has uncovered a network of more than 150 domains cloning law firms' websites, hidden behind Cloudflare and rotating IP ranges, in what SecurityWeek calls an AI-powered campaign. The sites target people already defrauded, offering to recover their losses with no upfront payment. Sygnia says it cannot yet tell how the scammers profit.
The production team behind Zero Day Attack has apparently brought a copyright complaint against the Kuomintang's new media team, alleging that AI-generated withered silvergrass in its satirical anti-recall ad copied green grass the producers filmed at great expense, according to KMT legislator Chen Ching-hui, who mocked the team's self-esteem as too fragile to survive an internet meme.
University of Valencia mathematician Pascual Diago used ChatGPT to write a nonsense paper on easing the anxiety of 60 pregnant women with mathematical metaphors, citing a fictitious Me-Lo I Nvent O, and the Clinical Journal of Obstetrics and Gynecology published it. The journal requested citations to four unrelated studies, and one "Robbie Williams" demanded $3,000. Diago answered with an AI-generated receipt from "Cheatbank from Spain".
An Office for Students investigation found that some staff at Regent College London encouraged business students to paraphrase content to lower plagiarism-checker scores, in some cases directed them towards AI tools or essay mills, and gave students early access to exam questions. The regulator placed the college under enhanced monitoring and imposed a new registration condition on partner University of Greater Manchester.
