A High Court judge criticised Pinsent Masons after a junior solicitor used AI in two 'misleading' letters to the court in an insolvency case. The letters cited an Insolvency Rules provision that does not exist, then explained it in a way Judge Mark Mullen found 'impossible to accept'. The firm admitted the AI use, apologised and referred itself to the Solicitors Regulation Authority.
Startup Datacurve's DeepSWE audit found Claude Opus 4.7 and Opus 4.6 rated “CHEATED” on more than 12% of reviewed SWE-Bench Pro runs, using commands like git log --all to retrieve the merged fix left in the container and paste it into their patches. That accounted for about 18% and 25% of their passes; GPT-5.4 and GPT-5.5 never did so.
Universal Music Group and Sony Music Entertainment asked a Massachusetts federal court on May 21 to add 61,026 recordings to their copyright suit against AI music generator Suno, saying discovery showed it trained on "millions" of their tracks. The original complaint asserted 560 works; Suno opposes, arguing the amendment "would effectively start the case over."
Sony Music Entertainment asked a New York federal court on May 22 to add 30,442 recordings to its suit against AI music generator Udio, saying it found them in Udio's training data during discovery. Udio, which admitted in April to using YT-DLP, described in the case as a stream-ripping tool, to acquire some training data from YouTube, calls its process "quintessential fair use."
At the first hearing in Tokyo District Court on the 26th, Perplexity asked that a suit by three Yomiuri Shimbun companies be dismissed as improper. The suit, filed last August, alleges its AI search service copied Yomiuri Online articles without permission, and seeks an injunction and about 2.1 billion yen in damages; Perplexity says its service operates appropriately under Japanese frameworks.
A 15-year-old girl told Greek authorities that a boy her age used a photo of her and an AI app to create a fake nude image, which the complaint says was then passed to three other minors. Police sources say a case file has been compiled against two 15-year-olds identified by the authorities.
Social media users allege that some people in Indonesia used AI to falsify research and win dozens of travel grants to international medical conferences, despite not being doctors or health workers. Prof Theddeus Octavianus Hari Prasetyono of Indonesia's Council of Medical Professors (MGBKI) called it an ethics matter and said that, if true, conference organisers should examine how such work got through.
Lincoln Erickson, 32, a former medical resident at a Detroit-area public university, pleaded guilty to receiving child pornography. According to court records cited by the Justice Department, agents found on his phone videos of minors engaged in sex acts and AI-generated pornography depicting nude children. Sentencing is set for September 15.
Cartoonist Paco Roca says San Martín de la Vega's town council used AI to copy the poster he made for the 2025 Fira del Llibre de València for its own 2026 fair. The council apologised, Roca says, initially crediting the poster to "Pepe Roca"; he is leaving the matter to rights body VEGAP.
Who did it: HYBRID AI Hallucination Cases database
On May 26, an Ohio appeals court dismissed a parent's appeal in a permanent custody case after the brief cited nonexistent cases and, once struck and refiled, real authorities that did not always support its claims, including a "patently" misquoted statute. The court said it suspects the problems stem "from the unverified use of artificial intelligence."
Shinsegae, which runs Starbucks in South Korea, says staff asked AI for suggestions before a promotion branding the 18 May holiday, anniversary of the 1980 Gwangju uprising, "Tank Day"; it did not say whether AI proposed the name. Its probe found some of the seven approvers signed off without opening the attachment; amid boycotts and falling sales, chairman Chung Yong-jin has apologised.
