Researcher Jeremiah Fowler found an unprotected database of South Korean firm GenNomis by AI-NOMIS holding 93,485 images and JSON files, vpnMentor reported, including explicit AI images of celebrities such as Ariana Grande, sometimes depicted as children. Public access was restricted after Fowler reported it. The company did not respond, and its websites later went offline with the database deleted.
US District Judge Dale Drozd in Sacramento let some plaintiffs pursue ERISA claims alleging Cigna used its PxDx algorithm to deny claims without medical director review. He called Cigna's reading that an algorithm may decide "so long as a medical director pushes the button" at odds with the plan's plain language. He dismissed wrongful-denial claims with leave to amend.
In Mavundla v MEC, the Pietermaritzburg High Court found that Mr Mavundla's lawyers cited many non-existent cases, likely AI-generated. When the judge consulted ChatGPT, it wrongly confirmed one of them. The court dismissed his bid to appeal and ordered the firm, Surendra Singh and Associates, to pay the costs of extra appearances. It also sent the judgment to the Legal Practice Council for investigation.
FTV's news channel used AI-generated footage circulating online in a 30 March report on the Myanmar earthquake, and netizens flagged it. FTV apologised on 31 March and pulled the story. The National Communications Commission said broadcasters must fact-check before airing, noting FTV had run corrections on its website and channel that day.
A YouTube channel called "Musk Chat" posted a deepfake of Elon Musk criticising the DPP's handling of the Yaya case. Former environmental chief Wei Kuo-yen then shared it on Facebook. The Foundation for Defense of Democracies had earlier flagged a similar channel, MuskTalk007, which uses AI to mimic Musk's voice and appearance to push pro-China, anti-US views.
An AI-made Facebook video shows NOVA presenter Ani Salich announcing that Assoc. Prof. Atanas Mangarov was shot outside his Sofia home, then advertises a supplement. A fake nova.bg article and a fake interview accompany it. Mangarov, very much alive, says he has repeatedly reported such scams to prosecutors and police, to little effect.
Bizly founder Ron Shah claims xAI's Grok chatbot infringes the "Grok" trademark Shah applied for in 2021. Shah says the startup's funding round collapsed after Musk's announcement and it now faces shutdown. Shah has not sent a formal demand letter. Separately, the USPTO has suspended xAI's own application over possible confusion with Groq and Grokstream.
Who did it: HYBRID AI Hallucination Cases database
The US Court of Federal Claims dismissed a pro se claim seeking over $1 billion in restitution from the United States over slavery, ruling it lacked jurisdiction. Judge Eleni M. Roumel found that cited cases such as "Fargo v. United States" do not exist and bear the hallmarks of AI. She warned that future fake citations may bring sanctions.
Who did it: HYBRID AI Hallucination Cases database
BC's Civil Resolution Tribunal ordered a former SQBox Solutions employee to pay $625 over a company laptop that was never returned. Tribunal member Douglas Merchant presumed the employee's submissions were partly or mostly AI-drafted: all five Employment Standards Act sections were misquoted, four cases did not support the stated propositions, and one did not exist.
