OpenAI will start watermarking ChatGPT’s text in the EU
techcrunch ·
OpenAI announced it will begin adding invisible watermarks to text generated by ChatGPT and Codex within the European Union to comply with the EU AI Act's transparency requirements. This technical measure, known as textGrain, subtly shapes word choices to create a detectable pattern without altering user experience or identifying individuals. The rollout is currently limited to the EU for end-users, while developers globally can enable the feature via the API on select models. OpenAI acknowledges that the watermark can be degraded through editing or synonym replacement, leading the company to restrict initial detector access to approved researchers. The firm cautioned that the absence of a watermark does not prove human authorship, as content may be too short, heavily edited, or sourced from other AI systems. This move follows similar commitments from competitors like Anthropic, although OpenAI previously delayed global implementation due to concerns about user retention. By adhering to these regulations, OpenAI aims to balance regulatory compliance with responsible AI deployment practices.
techcrunch