Oversight Board: AI Models Amplify Government Posture on Political Speech

Meta's Oversight Board tested 10 AI models from OpenAI, Google, Anthropic, xAI, and Meta using political criticism prompts. Models discouraged protest against speech-restrictive governments while encouraging it toward speech-permissive ones, mirroring each government's stance rather than applying a uniform standard. Co-chair Paolo Carozza called it "extended censorship by proxy that goes across borders."