Chatbots Still Bad at Spotting AI-Generated Images Despite New Law

Author Abdiaziz Ahmed details how some popular AI chatbots cannot reliably detect AI-generated images depicting false incidents of election-related malfeasance even after the implementation of the California AI Transparency Act. The law requires tech companies to embed origin data in AI-generated images, video, and audio; provide consumers with tools to access the origin data; and give users an option to include a visible indicator that content is AI-generated. After rerunning some tests from an earlier Brennan Center study, authors found that most of the chatbots didn’t mention the embedded origin data and still appeared to largely rely on often-faulty visual analyses of the images.

Go To Source