OpenAI will watermark ChatGPT outputs by default—but only in the EU
Summary
OpenAI will automatically watermark text generated with ChatGPT in the European Union, with the feature available by default there and optional in other regions. The watermarking technique, dubbed textGrain, aims to comply with the EU AI Act requiring detectable labeling of AI-produced content, but detection is not foolproof and can be circumvented. Competitor Anthropic has rolled out watermarking globally by default, while OpenAI’s approach is region-restricted for now.