A senior Trump administration official told Reuters that DeepSeek's latest model, due as soon as next week, was trained on Nvidia's Blackwell chips, which US export controls bar from China, likely clustered at its Inner Mongolia data center. The official would not say how DeepSeek got them. Nvidia declined to comment; DeepSeek did not respond.
The opinion of Minas Gerais appeals judge Magid Nauef Láuar, rapporteur in a February 11 majority ruling acquitting a man of raping a 12-year-old girl, contains on page 45 an AI prompt, 'Now improve the exposition and reasoning of this paragraph', followed by an AI-condensed version, Núcleo reports. The national justice council has asked the judge to explain the acquittal.
Google apologised after a Google News push alert about fallout from the Bafta ceremony, where an audience member with Tourette's syndrome used a racial slur in an involuntary tic, invited users to 'see more' on the slur. Google denied social media claims that generative AI was to blame, citing a failure in its notification safety features.
Johor police are investigating TikTok accounts that used AI to mimic the face and voice of state executive councillor Khairin Nisa Ismail promoting a non-existent Bantuan Kasih Johor aid scheme, which police say aimed to steal personal information. Police found 12 more accounts pushing the scheme and say no victims or losses have been reported so far.
The Upper Tribunal warned lawyers that putting client documents into ChatGPT breaches confidentiality and waives privilege after Manchester solicitor Tahir Mehmood Mohammed admitted uploading client emails and Home Office decision letters to it. He said he would inform clients, the IAA and the SRA; his appeal application had cited a fake case he guessed came from Google's AI mode.
Delhi High Court Justice Tushar Rao Gedela granted singer Jubin Nautiyal an ex parte interim injunction barring platforms, AI tools and online sellers from using his name, voice, likeness or singing style, including through AI voices and deepfakes, and ordered infringing links taken down. The suit alleges AI platforms were cloning his voice and imitating his singing.
Jason Powell, 31, of Bournemouth received a 15-month sentence suspended for two years at Bournemouth Crown Court after pleading guilty to making indecent photos of children and possessing prohibited images. Police found 93 AI-generated child abuse images after a sexual risk order flagged him downloading indecent images; the judge said he 'richly deserves immediate custody' but chose community treatment.
A Xi'an user who asked Tencent's Yuanbao app for a New Year's Eve greeting image built from his photo says that after about five prompts its 'Happy New Year' slogan was replaced by a four-character vulgar insult. As of February 23 his complaint had gone unanswered; in January Yuanbao apologised for earlier abusive replies, calling them rare anomalous model output.
A University of Crete political science professor says a student who failed for copying from AI skipped an offered oral exam, and his mother came instead, citing a stomach ailment and, with a colleague who travelled with her from Athens, pressing for a regrade while repeatedly invoking their professional roles. It was a first in his 23 years teaching there.
