Who did it: HUMAN Office of the Privacy Commissioner of Canada
Canada's federal privacy commissioner and its Quebec, British Columbia and Alberta counterparts found that OpenAI's collection of personal information from the internet to train GPT-3.5 and GPT-4 was overbroad and lacked valid consent, in contravention of privacy law. OpenAI disagreed with the findings. The federal office deemed the matter conditionally resolved after OpenAI's commitments; the BC and Alberta offices said the fixes fell short.
A Combat Antisemitism Movement report identifies 49 TikTok accounts in a coordinated network of fake AI-generated "rabbis" packaging antisemitic stereotypes as "insider Jewish financial wisdom", with a following of more than 950,000 users and over 10 million likes. CAM says a separate network of 70 fake rabbi accounts on Instagram was recently shut down with Meta.
Chad R. Snow, 42, a New York basketball trainer, was arraigned Wednesday on an indictment charging him with promoting and possessing a sexual performance by a child, obscenity and child endangerment over three AI-generated pornographic files that, the indictment indicates, all depict the same girl. The sheriff's office says he put a girl's face onto a pornographic video. He posted bail.
Who did it: HYBRID AI Hallucination Cases database
A federal magistrate judge sanctioned attorney Jessica Barsotti $1,000 after an opposition she filed in a suit against BART cited cases the court could not locate, three of four, raising concern about generative AI hallucinations. She failed to respond to an order to explain. She must take at least an hour of AI-ethics training; the order goes to her client and the State Bar.
After a scam call in March, a suburban Chicago man transferred $69,000 to two Wells Fargo accounts. A caller posing as a U.S. Marshal had texted him a photo of himself holding a badge bearing the name of a former Marshals Service deputy director. The Marshals Service told CBS News Chicago the photo was "completely generated by AI."
According to Dragos, an unknown adversary used Anthropic's Claude and OpenAI's GPT models in an intrusion into the IT network of a water and drainage utility serving Monterrey, Mexico. The firm says Claude independently flagged an industrial gateway as a target and ran a password spray against it. The spray failed, and Dragos saw no evidence that the operational environment was breached.
