Theverge OpenAI Introduces Watermarking for ChatGPT Text in EU to Comply with AI Act
Article Content
- •OpenAI's watermarking is designed to comply with the EU AI Act.
- •The watermarking feature is optional for API users outside the EU.
- •Editing text significantly reduces the watermark's detectability.
OpenAI is implementing invisible watermarking for text generated by ChatGPT and Codex in the European Union to comply with the EU AI Act, which mandates labeling AI-generated content. The watermark, known as textGrain, modifies word choices to embed a machine-readable signal that can be detected by approved tools. While the watermark will be mandatory for EU users, API customers globally can opt in to use it. OpenAI's internal tests indicate that the watermark is effective but can be diminished through text editing. The company emphasizes that the watermark does not verify authorship or accuracy and will not be a global default at launch. Access to detection tools will initially be limited to approved researchers and organizations. OpenAI plans to release the watermarking technology as open source in the future.
Ask AI about this cluster
Answers cite the sources they use