A Newsguard investigation reveals that leading chatbots struggle to identify AI-generated videos, raising concerns about misinformation. AI video generators like OpenAI’s Sora and Google’s Veo produce realistic footage that chatbots frequently misclassify. In tests, xAI’s Grok failed to detect 95% of AI-generated videos, while ChatGPT misidentified 92.5%. Google’s Gemini performed best at 78%. The study involved 20 doctored Sora videos with misleading claims, mimicking user queries. Despite watermarks indicating AI origin, Grok misclassified 30% and ChatGPT 7.5% of these marked videos. Furthermore, chatbots often provide confident yet incorrect assertions regarding authenticity, complicating misinformation detection. For instance, they accepted fabricated events as genuine news. Although Google’s Gemini can detect its own content, it lacks the capability to recognize outputs from other platforms. This highlights critical transparency gaps and the urgent need for improved AI literacy and detection methods in combating digital misinformation.
Source link
