Security researcher Alexander Hanff says installing Anthropic's Claude Desktop on macOS silently drops a Native Messaging manifest into multiple Chromium browsers' profiles, even for browsers not installed, pre-authorizing three extension IDs; it reappears on relaunch if deleted. Malwarebytes calls the 'spyware' label unfair but says it expands the attack surface without consent, and found no Anthropic rebuttal.
São Paulo civil police are investigating influencer Jefferson de Souza, 37, accused of using AI deepfakes to sexualize young Congregação Cristã do Brasil evangelical women and adolescents in videos posted without authorization. The probe followed a teenager's complaint in February. He admitted using churchgoers' photos and TikTok tools to animate them but denies the accusations, calling it humour.
Santa Catarina's Gaeco served search warrants in Itapema on Wednesday on suspects allegedly involved in using AI to imitate doctor Drauzio Varella's voice and image to promote illegitimate medicines on social media, supporting São Paulo prosecutors' Operation Double Check. Prosecutors say the drugs lack Anvisa registration; the case is under seal.
Security firm Expel says a North Korean state-sponsored group it calls HexagonalRodent used AI tools from OpenAI, Cursor and Anima to build malware and fake company websites, infecting more than 2,000 developers' computers via bogus job tests. Leaked wallet records suggest it may have stolen up to $12 million in crypto in three months; Cursor says it blocked the group.
The Consumer Federation of America has sued Meta, alleging it misled users and profited from fraudulent ads on Facebook and Instagram, such as fake 'free government iPhone' offers, some using AI to look credible, and charged high-risk advertisers more instead of banning them. Meta denies it, saying it removed hundreds of millions of scam ads last year.
Chile's Supreme Court suspended lawyer Daniela Beatriz Escobedo Acosta from practice for one month and imposed a fine of 5 monthly tax units after the opposing party doubted doctrinal citations in a cassation brief attributed to professors Juan Andrés Orrego Acuña and Jean Pierre Matus. The lawyer cited non-malicious transcription errors; the court said the sources went unverified.
Senator Maria Cantwell told HHS Secretary Robert F. Kennedy Jr. that WISeR, an AI-driven Medicare prior-authorization pilot launched January 1, 2026 in six states, is delaying care; a report from Cantwell's office using Washington State Hospital Association data says waits rose from about two weeks to four to eight. Kennedy called the delay 'unacceptable'.
An AI-generated deepfake image showing Westpac chief executive Catherine McGrath clashing with NZ First leader Winston Peters circulated on social media, clicking through to an investment scam. McGrath said Westpac tried contacting Meta via four routes and never got a response; the ad was eventually pulled, and the bank wants Meta to verify financial-services advertisers.
