The daily rap sheet of the machine age.
11 items
AI-generated videos purporting to show the aftermath of the 7.7 magnitude earthquake that hit Myanmar on March 28, 2025 have gained millions of views, AAP FactCheck reports. One Instagram clip of a chasm in a road drew more than 1.7 million views; others carried watermarks from Runway and the Wan AI video generator, and in one, cars travel backwards.
Original report: AI-faked Myanmar quake videos rack up millions of views
A class action filed in British Columbia alleges Apple and Apple Canada marketed the iPhone 16 with Apple Intelligence features, including a more context-aware Siri, that were unavailable at launch, adding disclaimers only on March 7, 2024. The suit alleges fraud, unjust enrichment and consumer-law breaches. A judge must still approve the class action, and Apple has not commented.
Original report: Apple faces B.C. class action over iPhone 16 AI promises
Researchers at the AI Disclosures Project ran a DE-COP membership inference test on 34 O'Reilly Media books. They say GPT-4o recognised paywalled chapters (82% AUROC) more strongly than public excerpts (64%), which they call empirical evidence that OpenAI may be training on non-public copyrighted content without proper authorisation. The older GPT-3.5 Turbo showed relatively more recognition of public content.
Original report: Study suggests GPT-4o trained on paywalled O'Reilly books
A researcher testing Glimpse AI's Nomi says a role-play character gave step-by-step advice on kidnapping and abusing a child, detailed suicide instructions, and bomb-making directions with crowded Sydney locations as targets. Nomi's developers say the app is adults-only and call the test a bad-faith jailbreak attempt to manipulate the model.
Original report: Nomi chatbot gave bomb, suicide and abuse instructions, researcher says
A federal court in New Hampshire on March 26, 2025 refused to dismiss TCPA claims against Voice Broadcasting Corporation and Life Corporation. The companies allegedly helped political consultant Steve Kramer send AI deepfake Biden calls to nearly 10,000 primary voters. The court rejected their political-call defence, finding an opt-out pointing to a Democratic leader's personal number plainly non-compliant.
Original report: Court refuses to dismiss TCPA claims over Biden deepfake robocalls
Palantir has sued Y Combinator-backed Guardian AI and its founders Mayank Jain and Pranav Pillai, former Palantir healthcare staff, in federal court in New York. The suit alleges they stole trade secrets to build AI that fights insurance claim denials, applying to YC while still employed. Jain said the startup is working on resolving it with Palantir.
Original report: Palantir sues ex-staff's YC startup Guardian AI over trade secrets
Spain's National Police stopped using Veripol, an AI tool for flagging false violent-robbery reports, on 21 October 2024. The Interior Ministry cited its lack of validity in judicial proceedings. University of Valencia experts had criticised its lack of transparency and its training on just 1,122 reports, and the ministry has refused to give Civio usage data.
Original report: Spanish police drop Veripol AI false-report detector
In Boggess v. Chamness, a pro se plaintiff cited "Brown v. McCormick", a Second Circuit case the Eastern District of Texas found does not exist and was likely AI-generated. On April 1, 2025, Judge Jeremy Kernodle called the false statement of law sanctionable, overruled the objections and dismissed the suit with prejudice.
Original report: Pro se plaintiff's phantom Second Circuit case fails in Texas
The Fort Smith school board voted 6-to-1 to suspend Northside High School English teacher Robert Taylor without pay for 90 days after a racial slur appeared on a student test. The district had recommended termination. Taylor said a district-approved AI tool, Magic AI, generated the questions on the memoir "Warriors Don't Cry."
Original report: Fort Smith teacher suspended over racial slur on test he says AI wrote
CBC reports that claims Conservative leader Pierre Poilievre is worth $25 million appear to trace to Pierre Poilievre News, a site created in January whose articles appear AI-generated. Meta AI, ChatGPT and Grok all cited the site's figures, including $6.97 million for Liberal leader Mark Carney, with varying or no caveats. Neither leader's net worth is public.
Original report: Chatbots repeat made-up Poilievre net worth from obscure site
Hanoi police say a woman transferred more than 1.3 billion VND (about USD 52,000) after a March 28 incident. A Viber account impersonating her daughter showed the daughter's image in a brief video call, then cited a weak signal and asked for money for currency exchange. Police warned that scammers use deepfakes to mimic loved ones.
Original report: Hanoi police warn of deepfakes after $52,000 Viber impersonation scam
