A 418-page, 10-year education reform plan for Newfoundland and Labrador, which recommends teaching AI ethics, contains at least 15 fabricated citations that academics suspect were AI-generated, CBC News reported, including a non-existent National Film Board film copied from a style guide's fictitious examples. The education department acknowledged "a small number of potential errors in citations" and said the report would be updated.
Who did it: HYBRID AI Hallucination Cases database
A California appeals court fined attorney Amir Mostafavi $10,000 after finding that 21 of 23 case quotations in his opening brief were fabricated. Mostafavi told the court he "enhanced" his drafts with ChatGPT and admitted he did not read them before filing. The court affirmed the judgment and directed that the opinion be sent to the State Bar.
The BJP lashed out at Congress over an AI-generated video in which a sleeping AI Narendra Modi dreams of his late mother Heeraben, who rebukes him for "doing politics in Bihar in my name." Union Minister Piyush Goyal cited Congress's "warped mentality"; Congress leader Pawan Khera asked where the disrespect was, insisting the clip offered "naseehat" (advice).
In a Sept. 11 letter to acting ICE director Todd Lyons, nine senators including Ed Markey, Jeff Merkley and Ron Wyden urged the agency to stop using Mobile Fortify, an app that lets agents scan faces and fingerprints for matches in government databases. They say it threatens privacy and free speech, especially for communities of color and protesters, and want answers by Oct. 2.
Turkish authorities have blocked access to the X account of Elon Musk's chatbot Grok, citing "national security and public order," the Freedom of Expression Association says, though the account remained reachable as of Sept. 12 with X yet to enforce the order. In July, an Ankara court restricted some Grok posts officials accused of insulting Atatürk and defaming President Erdoğan.
After the FBI posted blurry photos of the alleged killer of activist Charlie Kirk, users ran them through AI tools, producing dozens of "enhanced" images on X and web forums, many from Grok and others from ChatGPT, according to The Verge. The Hindu reports the chatbots also claimed the killer was a Utah Democrat, a hoax they attributed to CNN and the New York Times.
Who did it: HYBRID AI Hallucination Cases database
A federal judge in California ordered Robert Carey and Hagens Berman to pay $10,000, and co-counsel Celeste Boyd $3,000, after four opposition briefs in a suit against Fenix International contained hallucinated authority. Boyd said she relied on ChatGPT; the firm skipped a full cite-check. The December 12, 2025 order, which also refused corrective briefs, goes to the California and Arizona state bars.
The Medical Chamber of Slovenia says it has seen a rise in recent months in unknown parties using prominent doctors' names, photos and AI-altered video statements, without consent, to promote "miracle cures." It reports the cases to market and health inspectors, the medicines agency JAZMP and police, but says the fakes spread ever faster, often via foreign servers and paid social-media ads.
THORChain co-founder JP lost around $1.3 million in crypto after attackers used a hacked Telegram account and a deepfake video call to trick him into merging his call, then accessed his encrypted iCloud and keychain and drained a forgotten MetaMask wallet, Coinspeaker reports, blaming North Korean hackers. His multi-signature Vultisig wallets were untouched, he said.
Who did it: HYBRID AI Hallucination Cases database
A federal judge in California ordered Luther & Lanard, PC to pay opposing counsel $2,418.50, half the fees spent addressing false citations in a July 14, 2025 brief that quoted nonexistent language and cited nonexistent cases. The firm said its AI produced them and the drafting attorney did not verify them. It must also notify the California State Bar.
