Microsoft's threat intelligence team says two North Korea-linked groups, Jasper Sleet and Coral Sleet, use AI to research jobs, write résumés, invent personas and headshots, face-swap stolen identity documents and use voice-changing software in interviews. Once hired, they keep up the act by having AI help answer emails and write code snippets. Coral Sleet also uses AI to build malware and infrastructure, the report says.
Hachette, Penguin Random House, HarperCollins, Macmillan, Simon and Schuster and several academic publishers sued "shadow library" Anna's Archive in New York federal court, alleging it pirated over 63 million books and provides them to companies for AI training. They seek a shutdown and up to $150,000 per pirated work; the site, run by unknown individuals, has not responded to a related Spotify suit.
Magistrate Judge Robert Numbers said in a March 2 order that a brief signed by assistant US attorney Rudy Renfer contained "fabricated quotations and misstatements of case holdings," and ordered leaders of the Eastern District of North Carolina US attorney's office to show cause why sanctions shouldn't follow. Renfer blamed an unfinalized draft; whether AI was used is not yet clear.
On El Hormiguero, host Pablo Motos aired an AI-generated video of a fake Motos promising viewers 19,000 euros in a month, saying it is false and that victims of such scams "lose everything", some in Spain over a million euros. He accused Facebook, Instagram and YouTube of being "complicit" for not cooperating despite years of work with Spain's National Police.
Jonathan Bates was sentenced to five years in prison for stalking offences after creating fake pornographic accounts for two women he had worked with, reportedly putting their faces on naked bodies. One woman says a man who believed he was in a relationship with her came to her home after Bates gave him her address, and that she has since quit her job.
Justice S Mohan ordered lawyers Goh Peck San and Amarjit Singh Sidhu to pay S$5,000 each in personal costs after two fictitious cases, which the judge said most likely came from AI, appeared in closing submissions. Mr Singh blamed a since-departed paralegal; the judge called their S$1,500 offer "plainly inadequate" and found it "extraordinary" the issue "fell between two stools."
Who did it: HYBRID AI Hallucination Cases database
US District Judge Jerry Edwards Jr. personally sanctioned lawyer Kenneth Mayeaux $1,000 after his ChatGPT-drafted memorandum, supporting a motion to block a petitioner's removal, cited nonexistent cases and incorrect quotations. Mayeaux, who admitted not fully verifying the citations, had proposed the $1,000 figure himself; the judge commended his candor but said Rule 11 requires attorneys to verify submissions.
Grammarly's 'expert review' feature offers AI writing suggestions 'inspired by' real experts, including Verge editor-in-chief Nilay Patel and other Verge staff, none of whom gave permission, The Verge reports. Parent company Superhuman says the agent 'doesn't claim endorsement' and that experts appear because their published works are 'publicly available and widely cited'; some cited 'sources' led to spammy copies or unrelated pages.
A US unit of Nippon Life Insurance has sued OpenAI in federal court in Chicago, alleging ChatGPT gave unlicensed legal advice to a disability policyholder who, after settling a lawsuit against the insurer in 2024, used the chatbot to build arguments and draft documents to revive the case. The insurer seeks damages for the time and money spent.
