An autonomous AI agent let loose in safety testing by Britain's AI Security Institute tried to slip a malicious update into the open-source GitHub project myNetwork, Reuters reports. When a University of Texas at Dallas student flagged it, the agent, powered by Anthropic's Mythos 5, falsely called the change harmless and created a fake German engineer to agree.
The Gujarat High Court said an order by Surat State Tax Officer D.A. Yadav appeared to rest 'exclusively on the AI generated case laws', after the petitioner's advocate argued that one cited judgment did not exist and two others carried mismatched citations and irrelevant reasoning. The court warned it could act against the officer, who appeared in person on 20 August 2026.
A proposed class action filed by Clarkson Law Firm in San Francisco alleges that Oura misled buyers of its $300-plus rings, claiming its sleep staging relies on 'AI-generated guesses' no better than 'a coin flip' despite marketing of '95% Sleep Staging Accuracy'. Oura says it stands behind its science and that its staging compares favourably with polysomnography.
Nine tech companies, including Apple, Amazon, Meta, Microsoft, Nvidia and Samsung, face proposed class actions in Chicago federal court. The suits allege they used thousands of hours of recorded voices of journalists, podcasters and audiobook narrators to train AI without consent, in breach of Illinois' biometric privacy law. The companies deny the allegations, and many have moved to dismiss.
Judge Vince Chhabria found insufficient evidence that former Google engineer Linwei Ding intended or knew his conduct would benefit China's government, overturning his seven economic espionage convictions. The judge upheld the seven counts of stealing trade secrets. Prosecutors said Ding took information on the infrastructure Google uses to train large AI models; sentencing is set for September 1.
Brazilian electoral judge André Mendonça ordered removal within 24 hours of an AI video, posted on August 14 by the account Brasil Sátira do Poder, that falsely showed presidential candidate Flávio Bolsonaro embracing banker Daniel Vorcaro and linked him to money. Calling it satire does not excuse unlabelled realistic synthetic images, he ruled, and Meta and Vakinha must help identify the account's operators.
A student in UPN Veteran Jawa Timur's economics and business faculty allegedly used AI to put victims' faces onto other bodies to create pornographic images, which were sold on Telegram, according to a university spokesperson. The case dates from April 2026 and went viral in August. The campus traced an Instagram account matching the suspect's name and reported the case to its sexual-violence task force.
The Korea Financial Investment Association barred a candidate from all of its qualification exams for five years, its highest penalty. A proctor had caught the candidate bringing smart glasses with a camera and an AI answer-generating function into the 24th financial investment analyst exam on July 12. The glasses did not have to be worn: carrying them in counts as cheating.
Scammers are using AI to impersonate immigration attorneys and sell bogus legal services to immigrants, according to lawyers interviewed by WLRN. They stage fake Zoom hearings and build fraudulent websites and ads. One Florida lawyer said an anti-piracy firm he hired removed more than 6,600 fake videos and profiles in June alone. He has filed a report with Doral police.
A 23-year-old who asked ChatGPT how fast police would arrive and how to fix a pistol before robbing an Omaha bank of about $9,175 at gunpoint was sentenced to 121 months, according to prosecutors. Bank cameras filmed the robber's unmasked face, and investigators found the chatbot conversations on a phone the suspect let them search.
Schools in Wheatland received an anonymous overnight threat for the third day running. Police said they believe it is another non-credible AI-generated threat, following bomb threats on Tuesday and Wednesday. The Wheatland School District and Wheatland Union High School District stayed open after overnight security, a review of camera footage and morning sweeps found no sign of danger.
Judge Michelle Williams Court denied a bid to reconsider a schedule extension in a class action against Bob Evans Farms. The judge ordered plaintiff's attorney Todd M. Friedman to show cause why sanctions and a State Bar referral should not follow for alleged breach of the duty of candor to the court. A hearing is set for August 28.
