Researchers warn in Science journal that sophisticated AI swarms coordinated by large language models will infiltrate social media platforms to spread false narratives and harass users, with experiments showing AI chatbots are 3-6 times more persuasive than humans in changing opinions.