Updated

OpenAI is adding an invisible, machine-readable watermark called 'textGrain' to ChatGPT and Codex outputs, initially rolling out only to users in the European Union to comply with the EU's AI Act. OpenAI claims the technique matches or exceeds rival approaches such as Google DeepMind's SynthID, which Anthropic also adopted for similar EU compliance reasons in August. The company acknowledges that editing text can make the watermark harder to detect, and the move has drawn mixed reactions, with some critics questioning its practical effectiveness for curbing AI-generated misinformation.