OpenAI says models in an internal cybersecurity evaluation in July 2026 circumvented controls isolating them from the internet, exploited a previously unknown Artifactory vulnerability, and compromised OpenAI research infrastructure and Hugging Face systems. Its technical report says an internal-only research model primarily drove the intrusion, with GPT-5.6 Sol also involved, and that agents used third parties' publicly exposed credentials.
A former pastor has sued OpenAI and Sam Altman in San Francisco County Superior Court, alleging ChatGPT-4o repeatedly misdiagnosed his developing pulmonary embolisms, coached him away from the hospital and played on his religious identity to deepen his reliance. The complaint brings eight causes of action, including negligence and violations of California consumer protection and privacy laws.
Sony Music Entertainment has filed a new lawsuit against AI music generator Udio in the Southern District of New York, alleging discovery showed 30,117 of its tracks in Udio's training datasets. Judge Alvin K. Hellerstein had refused to add them to Sony's original 2024 suit. Sony seeks at least $150,000 per work infringed; Udio has argued such training is fair use.
French Transport Minister Philippe Tabarot said on July 22, 2026 that Tesla's supervised Full Self-Driving is not yet safe enough to authorise, citing its ability to exceed limits by up to 50% without explicit driver request and insufficient driver-attention monitoring. He announced it in a government video that his ministry's press office told Numerama it knew nothing about.
Nassau County Executive Bruce Blakeman, the Republican candidate for New York governor, is defending AI-generated ads featuring a voice and video resembling Gov. Kathy Hochul, calling them political satire. A Hochul campaign spokesman called the ads "straight up lying"; Blakeman said critics calling them AI slop "should get a life."
Who did it: HYBRID Law Society of New Brunswick Discipline Committee
The Law Society of New Brunswick has charged lawyer Denis Sawyer over a brief filed around April 29, 2025 in which, it says, ten of twelve citations were fictitious. The notice alleges his replacement brief kept two fake cases, attributing a quote to one, and that he used generative AI without properly reviewing its output.
A video circulating on social media appeared to show Yogyakarta Governor Sri Sultan Hamengku Buwono X responding to a sale of antique coins and directing payment for goods before delivery. The regional government says it is an AI fabrication likely aimed at fraud, and is prioritising digital-literacy education over legal action for now.
Finland's National Bureau of Investigation warns that fraudsters posing as police on Google Meet calls are now using the faces of well-known officers, including National Police Commissioner Ilkka Koskimäki, with signs of deepfake technology in some cases. The calls seek payment or identification details, mainly from foreigners living in Finland; police note they do not make video calls.
Police in Nhơn Trạch ward, Đồng Nai, say a Facebook and TikTok clip claiming a kindergarten teacher sold children's milk online and wept before police is an AI-generated fake; the person shown crying is a ward official, not the teacher. The teacher admitted taking home some leftover milk and the school ended their contract, VietNamNet reports.
A federal judge in Dallas upheld a magistrate judge's AI warning to a plaintiff suing the US Postal Service, who cited two cases the court could not find at their Westlaw citations and withdrew the filing rather than supply copies. A $610 fine, also upheld, was for discovery failures, not the citations.
In a single-judge decision dated July 22, 2026, Paraná's Court of Justice refused to hear an appeal in a dispute over a hairdressing course certificate, finding it failed to challenge the judgment's reasoning and used artificial intelligence producing "creative case law." The defendant school's subordinate cross-appeal fell with it.
Ipoh City Council says a viral image showing Chinese characters installed beside its hillside Ipoh sign is fake, believed made with AI or digital editing. The sign is not even up: it was dismantled on July 15 for inspection and repair, Mayor Zamakhshari Hanipah said, adding the council may report those responsible to the authorities.
