DREAM Lab says a framework built on Hermes and OpenClaw agents, running up to eight sub-agents in parallel, spent July 1-4, 2026 attacking unnamed Asian government entities, cracking 85 credentials and exfiltrating personnel records. Model refusals were bypassed by framing everything as "authorized penetration testing"; the researchers say linguistic clues point to a Chinese-language operator.
A 17-year-old in Massachusetts was charged with two counts of murder after his mother and 14-year-old brother were found dead at home. Middlesex District Attorney Marian Ryan said he had used ChatGPT to create "fantasy stories, Gothic novels" which "appear to relate to threats to his family not surviving".
Berlin non-profit HateAid has filed a criminal complaint against Meta, Ray-Ban, Oakley and several retailers, arguing that selling Meta's smart glasses in Germany breaches the Telecommunications, Digital Services and Data Protection Act because wearers can film people covertly. Meta says the glasses were approved by regulator BNetzA in 2022 and have a tamper-resistant recording LED.
Twitch announced on Wednesday an opt-out setting to stop Amazon's AI models training on users' content, which doubled as confirmation the training was happening; users are opted in by default. Chief product officer Mike Minton said he didn't know what Amazon had used, adding, "I'm not responsible for Amazon's training efforts."
A 7amleh and Utrecht University report on Meta's alleged suppression of Palestinian voices analysed 3,520 user-logged cases from 2021 to 2025, finding only 32.5 percent had an identifiable policy violation and 57.2 percent of those cited the Dangerous Organizations and Individuals policy. It calls the pattern structural, not error alone, while stopping short of asserting deliberate censorship.
New Bedford officer Emily Pacheco was placed on paid leave on Monday, the day her ex-girlfriend obtained an abuse prevention order; the woman's complaint says internal affairs is investigating Pacheco for running her plate through Flock to follow her. Pacheco faces no criminal charges, and the department has paused its Flock cameras after an audit.
Florida's Fourth District Court of Appeal referred lawyer Barry M. Leff to the Florida Bar over a certiorari petition "riddled with false citations," including a non-existent case attributed to the court itself. Leff said he filed the AI's initial draft by mistake; the court noted the version he said he meant to file also omitted controlling law, settled since 1979.
A federal judge in the Northern District of Alabama sanctioned attorney Scott Thomas Morro after his brief opposing dismissal cited two cases that do not appear to exist and attributed quotations to seven cases that do not contain them. A Westlaw editor's note says the citations were likely AI-generated; Morro's earlier correction notice withdrew only one fabricated case.
Pernambuco's Regional Electoral Court ordered immediate removal of an Instagram post by @juventudejoaocampos showing an AI-simulated photo of Governor Raquel Lyra beside presidential candidate Flávio Bolsonaro, captioned "In Pernambuco, Bolsonarism is purple". Judge Fernando Braga Damasceno faulted its lack of AI labelling, which the court says breaches TSE rules, and ordered Instagram to identify the account holder.
The Nova Scotia Court of Appeal dismissed self-represented appellant Kevin Joel Arbuckle's appeal after he admitted using an AI platform, not a certified court reporter, to produce the transcripts in his appeal book, which nonetheless carried a certification. He admitted it only after the purported transcriber denied the work, and blamed the false certifications on AI hallucinations.
