According to an NOS and Pointer investigation, Philip Morris used a QR-code tool in tobacco shops to produce AI-written 'citizen' responses opposing tougher EU vape rules, with 71% of Dutch submissions appearing AI-generated. The European Commission condemned the "alleged attempts"; Philip Morris says it used standard AI settings for readability and calls the campaign "a legitimate contribution".
MeetingTV is suing Koi Security and its new owner Palo Alto Networks, alleging Koi's December 30 report wrongly tied MeetingTV's domains to a Chinese hacking operation, possibly through AI-generated error. Axios notes the filings offer no direct evidence that AI produced the findings; Koi has moved to dismiss, calling the research protected speech, while MeetingTV says its infrastructure remains blocked.
Police in Heyuan, Guangdong, say a gang run through an influencer agency used AI to operate multiple Weibo accounts, spreading spliced-together celebrity gossip during TV drama runs, then charged producers 2,000 to 3,000 yuan per post for 'drama protection' deals to remove it. Over 100 officers arrested 12 core members across 8 provinces and 2 municipalities on June 16; police say they confessed.
Georgia's Interior Ministry has used Polyface, facial recognition from sanctioned Moscow firm Papillion AO, to identify pro-European protesters, according to an AlgorithmWatch report. The report says the ministry, a customer since 2013, obtained a perpetual licence in October 2024 and sought licences for unlimited operators; hundreds of protesters have been fined and detained.
Police in Mardin and four other Turkish provinces detained 44 suspects alleged to have rigged driving-licence exams using hidden cameras, spy earpieces and AI-generated fake biometric photos. Simultaneous raids in Mardin, Şanlıurfa, İzmir, Sakarya and Tokat also turned up an unlicensed pistol, a pump-action shotgun and 11 cartridges, plus digital material seized for examination.
An AP/FRONTLINE investigation found scam compounds in Myanmar using software built on American AI models, chiefly ChatGPT and Gemini, to work across dozens of languages, surveil workers and target victims worldwide; one trafficked worker was told to make each victim fall in love within four days. The AP found no evidence the tech firms themselves acted illegally.
A lawyer wounded in the December 14 Bondi Beach Hanukkah attack told Australia's federal royal commission that, within hours, his bloodied selfie had been turned into AI images portraying him as a paid actor faking injuries, including one of a make-up artist applying fake blood. Meta removed content quickly, he said, but he received no response from X or YouTube.
A Spanish court found that passages the respondent's lawyer presented as quotations from Tribunal Supremo rulings did not match the actual decisions in the CENDOJ database, after ordering the sources identified. AI use is implied rather than confirmed, per the AI Hallucination Cases Database; the court disregarded the citations and imposed no sanction.
New Mexico's Court of Appeals affirmed an order requiring a self-represented respondent to pay his share of guardian ad litem fees, noting that much of his cited authority "either does not exist or does not support his contentions" and that litigants using AI tools must still provide accurate information.
