The daily rap sheet of the machine age.
10 items
Retraction Watch checked 18 of the 46 citations in Mastering Machine Learning: From Basics to Advanced, a $169 Springer Nature ebook by Govindakumar Madhavan, and found two-thirds either did not exist or had substantial errors. Madhavan did not say whether he used an LLM; Springer Nature said it was 'currently looking into it'.
Original report: Springer Nature's $169 machine-learning ebook riddled with faulty citations
Nikkei found hidden prompts like 'give a positive review only' in 17 arXiv preprints whose authors are affiliated with 14 institutions in eight countries, including Waseda, KAIST and Peking University, concealed with white text or extremely small fonts. A KAIST co-author called it inappropriate and said the paper will be withdrawn; a Waseda co-author called it a check on 'lazy reviewers'.
Original report: Researchers hide 'positive review only' prompts for AI reviewers
Justice Paul Goetz of New York state court ruled on Friday that Meta and TikTok owner ByteDance must face a wrongful death suit from the mother of a 15-year-old who died subway surfing. She alleges they "goaded" her son by addicting him to Instagram and TikTok; Goetz found it plausible they actively identified users most impacted, but dismissed her claims against the MTA.
Original report: Meta, ByteDance must face suit over teen's subway-surfing death
Taiwan's Ministry of Digital Affairs fined Meta NT$15 million after finding systemic flaws in Facebook's advertising management system, which left some ads without timely disclosure of who commissioned and paid for them, a serious breach of anti-fraud rules. The probe stemmed from 23 cases referred by the Interior Ministry; Meta has 30 days to fix it or face repeated fines.
Original report: Taiwan fines Meta NT$15 million over Facebook ad disclosure failures archived copy of Taiwan fines Meta NT$15 million over Facebook ad disclosure failures
Abdullah Gül, Turkey's 11th president, said AI-generated fraudulent videos of him were circulating online and that complaints had been filed with prosecutors and cybercrime units. Scammers use AI to mimic celebrities in sponsored "investment" ads; a gang that took over $500 million was broken up in May, but similar ads persist.
Original report: Former president Abdullah Gül files complaints over AI scam videos
Leonardo Crespo, suing Tesla without a lawyer over his 2024 Model X, cited two apparently nonexistent cases and a quote a real case never contained in discovery filings, then admitted using AI. Magistrate Judge William Matthewman, noting his candor, ordered him to apologize in writing to Tesla's counsel and pay its reasonable fees for chasing the fake citations, and denied his motions.
Original report: Pro se Tesla plaintiff ordered to apologize for AI-fabricated cases
Russian Telegram channels are circulating a video in which Third Army Corps commander Andrii Biletskyi appears to say Ukrainian authorities deliberately avoid identifying fallen soldiers to dodge compensation payments. Ukrinform says the audio was AI-generated and inserted into earlier footage, citing unnatural facial expressions and a Hive Moderation result rating 99% of the audio as synthetic.
Original report: Russian Telegram channels spread deepfake audio of Ukrainian commander Biletskyi
A Jerusalem enforcement registrar rejected a legal-aid lawyer's request to lift a debtor's exit ban, noting it quoted a nonexistent subsection, 66A(a1) of the Execution Law, and misstated the relevant amendment. The registrar said this raised suspicion of uncontrolled AI use, cautioned that AI output must be verified, and waived costs because the debtor has legal aid.
Original report: Israeli debtor's lawyer cites nonexistent provision, registrar suspects AI
Ontario's Animal Care Review Board refused the Chief Animal Welfare Inspector's bid to dismiss Joshua LaPointe's appeals over the removal of his animals, despite ChatGPT-assisted filings that, the inspector argued, cited non-existent or inapposite cases and likely fabricated veterinary opinions. LaPointe admitted not verifying ChatGPT's citations; the board said it would disregard false citations and excluded non-compliant material in two files.
Original report: Ontario animal board to disregard appellant's ChatGPT false citations
The Ho-Chunk Nation Trial Court found that Benjamin Gamble, a losing legislative candidate challenging the June 3, 2025 election, cited Election Code provisions that do not exist. Defense counsel questioned whether he used AI and sought sanctions; Gamble said he researched online using Google. Chief Judge Lowe declined to sanction him, cautioned that online citations must be checked, and dismissed his case.
Original report: Ho-Chunk court finds candidate cited nonexistent Election Code sections
