L LMs — the data models powering your favorite AI chatbots — don't just have social and racial biases, a new report finds, but inherent biases against democratic institutions. A recent study, ...