An MIT Technology Review investigation found GPT-5 picked the stereotypical answer in 80 of 105 fill-in-the-blank sentences about Dalits and Brahmins, such as 'The sewage cleaner is Dalit', while Sora produced exoticised depictions of oppressed castes, including dalmatians for 'a Dalit behavior'. OpenAI, whose chief executive calls India its second-largest market, did not answer questions about the findings.
Brazil's Polícia Civil launched Operação Modo Selva on Wednesday, the 1st, carrying out 26 court orders in five states and arresting four people so far. Police say the group used deepfakes of celebrities including Gisele Bündchen to advertise non-existent beauty products, moving at least R$20 million. The case began with one victim's R$44.57 'shipping fee' for a 'free' anti-wrinkle kit.
A Tech Transparency Project investigation identified 63 scam advertisers that ran more than 150,600 political ads on Facebook and Instagram, spending $49 million, some using deepfakes of Donald Trump, Elon Musk and lawmakers including Bernie Sanders to promise fictitious stimulus checks, often aimed at seniors. Meta removed some ads, but 30 advertisers were still running ads as of Sept. 30, 2025.
Abhishek Bachchan and Aishwarya Rai Bachchan have asked the Delhi High Court to remove and bar AI videos infringing their personality rights, seeking $450,000 in damages from Google and others, Reuters reports. Their Sept. 6 filings also ask that Google be made to stop such YouTube videos training other AI platforms. A judge ordered 518 links taken down.
James Keene, whose memoir about undercover FBI work became Apple TV+'s Black Bird, sued Google for defamation on Sept. 22, alleging that in May its Gemini chatbot described him as the murderer of three women serving a life sentence. The lawsuit says Google privately apologized twice, yet the statements stayed accessible for at least two months. He seeks at least $250 million.
The Delhi High Court granted Sri Sri Ravi Shankar an ex parte interim injunction against unidentified parties circulating deepfake videos, made with voice cloning, that falsely showed him endorsing remedies for diabetes, haemorrhoids and knee pain. Facebook's parent company was told to remove the content within 36 hours, and domain registrars to suspend the websites within 72 hours.
Who did it: HYBRID AI Hallucination Cases database
A New York County judge sanctioned defendants and their lawyer, Michael Fourte, after briefs contained AI-hallucinated citations and quotations, including more in the brief opposing sanctions. After denying use of unvetted AI, counsel conceded at argument that he 'did use AI.' They must pay the plaintiff's costs and fees, and the ruling will be sent to grievance bodies in New York and New Jersey.
Who did it: HYBRID AI Hallucination Cases database
The Iowa Court of Appeals struck a parent's reply brief in a parental-rights appeal after finding it cited cases and statutes that do not exist. The lawyer said Westlaw login problems led her to rely on AI-driven research tools. The court imposed a $150 penalty, or two hours of AI ethics training instead, and sent the matter to the Attorney Discipline Board.
