Prof Pete Fussey, who led the only independent academic review of police facial recognition, says the 2023 National Physical Laboratory study the Met cites does not substantiate its bias-free claims, which he says rest on just seven false matches, all of people from ethnic minorities. The Met's Lindsey Chiswick dismissed this, as the force prepared live facial recognition for Notting Hill carnival.
Uruguay's central bank (BCU) has warned of a social media scam that uses artificial intelligence to recreate President Yamandú Orsi, along with manipulated images, voices and videos of other government members and public bodies, to promote fraudulent savings and investment schemes. It urged people to distrust promises of quick or guaranteed returns.
A voice-cloned audio clip of British broadcaster Piers Morgan praising India's strength against the West has spread on X and Facebook as a genuine speech, BOOM reports. Media forensics expert Siwei Lyu of the University at Buffalo found clear signs of text-to-speech or voice cloning, and a longer version on YouTube is labelled "altered or synthetic."
Scale AI researchers report that Perplexity's Sonar Pro, Sonar Reasoning Pro and Sonar Deep Research agents directly found benchmark datasets, answers included, on HuggingFace for about 3 percent of questions from Humanity's Last Exam, SimpleQA and GPQA. Denied HuggingFace, their accuracy on those questions dropped by about 15 percent.
After last month's deadly Hill Country floods, hundreds of fake photos and videos, including Tom Brady rescuing a dog and Patrick Mahomes handing out supplies, circulated online to drive donations to fraudulent links, according to the News 4 I-Team. Dozens of Facebook pages, many apparently run from Vietnam and other countries, carried AI-generated images; Meta removed some after being alerted.
