Hackers hijacked Instagram accounts, including the Obama-era White House handle, by asking Meta AI Support Assistant to add a new email to a target's account, according to a video posted on X; the chatbot sent a code to that address and then offered a password reset. Instagram spokesperson Andy Stone said Monday the issue was fixed; how many accounts were hit is unclear.
Reporters found AI-generated "doctors" on TikTok, YouTube and Facebook promoting unproven home remedies, such as potato juice for stomach ailments and garlic, lemon and honey for cataracts; six TikTok health videos alone totalled over 14 million views. The videos funnel audiences to Telegram groups selling "miracle" products, the report says. TikTok removed some profiles after reporters' questions.
Who did it: HYBRID Florida Office of the Attorney General
Florida Attorney General James Uthmeier sued OpenAI and CEO Sam Altman in what he calls the first state-led suit of its kind. The complaint alleges they marketed ChatGPT to the public, including children, while concealing risks, ignoring safety warnings and deploying a product that encourages self-harm and violence. A separate criminal investigation, opened after prosecutors reviewed the Florida State University gunman's ChatGPT chats, is ongoing.
Vanderbilt researchers worked from a trove of 100,000 Geedge Networks documents. They say the Chinese company, which sells a commercial version of the Great Firewall, has been developing AI that profiles citizens from telecommunications, social media and location data to highlight who may pose a political risk. The work appears to be at the research stage, its progress apparently hampered by US chip export controls.
Users of the second-hand marketplace Xianyu said it appeared to turn photos on their phones into sale listings without their knowledge. The listings covered their collectibles, pets and even museum artifacts, and some users only noticed when buyers got in touch. On June 1 Xianyu said its AI identified the images as collectibles and generated suggested titles and descriptions. It apologised, promising stronger listing confirmation.
Who did it: HYBRID AI Hallucination Cases database
The Seventh Circuit fined attorney Abdullah Salah $5,000 after an immigration brief he admitted signing without review contained 24 unattributable or false quotes and at least seven mislabeled or nonexistent cases. Its author, Farah Chalisa, said ChatGPT inserted the errors during a grammar check. The court found that explanation wanting, admonished her, and referred the matter to Illinois's disciplinary commission.
Chad R. Snow, 42, a basketball coach from Camillus, N.Y., appeared in court Monday facing 14 charges. The charges relate to AI-generated pornographic images the Onondaga County Sheriff's Office says it found on his device. Victims were aged 14 to 30, including a girl who was 12 in the original photo. Both sides have substantial discovery to review, and his next hearing is July 20.
French TV host Michel Drucker warned his Instagram followers that a fake account was using his image and voice to sell a joint-pain medicine. The outlet counts it among many celebrity impersonations it links to more accessible AI. 'Don't get caught out,' he wrote. He named the site, Health Bridge, and dismissed the product as 'poudre de perlimpinpin', or snake oil.
A São Paulo judge issued a preliminary order for Facebook to remove posts using AI-generated images. Senator Magno Malta's lawyers say the images pushed 'false narratives' about an episode in which he was accused of assaulting a nursing technician. Judge Miguel Ferrari Júnior also ordered the users' access records and registration data preserved. Malta denies the assault accusation, which is under investigation.
Police in Palembang's Gandus district detained a 24-year-old identified by the initials F alias E. Police say F admitted that a viral photo of a pocong, a shrouded ghost, was made with AI. Many residents believed the image was real and shared it widely in chat groups, spreading panic. Police are still examining the motive and have not yet named F a suspect.
Who did it: HYBRID AI Hallucination Cases database
A Queens judge struck pro se defendant Mendel Hendel's answer in a rent dispute and found him in default. It held his filings amounted to fraud on the court, citing the nonexistent 'Ennis v. Lessing', quotes absent from real cases and a bare removal notice. The judge said the errors looked AI-generated, which Hendel denies, and ordered him to reimburse part of the plaintiff's fees.
In a mortgage enforcement case, Ontario Superior Court Justice Stewart cautioned a defendant and his lawyer over a factum served under counsel's name, which counsel said the client had prepared. It cited 'Waldick v. Penney, 2019 ONCA 201', but no such case could be produced and the citation belongs to an unrelated case. The judge denied the defendant costs, partly over the misleading citation.
