South Korea's Northern Gyeonggi Provincial Police Agency, working with HYBE under a February 2024 agreement, arrested eight people suspected of creating and distributing deepfake videos of HYBE Music Group artists. HYBE supplied information and fans sent tips; CEO Jason Jaesang Lee promised a 'zero-tolerance and no-settlement policy'. Neither the suspects nor the affected artists were named.
Northern Gyeonggi police said they caught about 100 men in their teens to 30s over sexual deepfakes of some 100 victims, including female idol group members, actors and internet streamers. Police say 23 Telegram chat room operators were caught, 13 referred to prosecutors in custody; one allegedly made over 1,100 videos, another used 'deepvoice' to fake vulgar speech.
Twelve former OpenAI employees, in a proposed amicus brief filed by Lawrence Lessig in Elon Musk's lawsuit, argue that ending the nonprofit's control would "fundamentally violate its mission" and breach staff trust, saying the structure was used to recruit and retain them. OpenAI says its nonprofit "isn't going anywhere"; a jury trial is set for spring 2026.
Ireland's Data Protection Commission has opened an inquiry into whether X lawfully processed publicly accessible posts by EU/EEA users to train its Grok AI models, examining GDPR lawfulness and transparency. The DPC, X's lead EU regulator, can fine up to 4% of global revenue; X previously agreed to stop using EU user data for AI training without explicit consent.
A group of copyright law professors filed an amicus brief backing authors including Richard Kadrey, Sarah Silverman and Ta-Nehisi Coates, who allege Meta trained its Llama models on their e-books without permission, calling Meta's fair use defence "a breathtaking request for greater legal privileges." Publishers' groups also backed the authors; the Electronic Frontier Foundation and other professors sided with Meta.
Who did it: HYBRID AI Hallucination Cases database
Paraná's Court of Justice unanimously refused to hear a murder defendant's appeal after finding all 43 precedents it cited were AI inventions, from judges who do not exist to case numbers like "1234-56", and that quotes attributed to the lower-court ruling were not in it. Lawyer Roberley Elias was warned and no fees were awarded.
