OpenAI to Add Digital Watermark to Text and Code Generated in the European Union
AI developer OpenAI plans to embed an invisible digital watermark in text and software code generated in the European Union. The move is meant to bring the company in line with the EU's new AI transparency rules.
AI developer OpenAI has announced that it is preparing to comply with the European Union's new transparency requirements for artificial intelligence.
The company will soon begin attaching an invisible digital watermark to text and software code that its models generate within the EU. The watermark cannot be seen by the user, but it makes it possible to identify content as produced by artificial intelligence.
A similar step in the industry
OpenAI's decision follows a comparable initiative introduced earlier this year by another major industry player, Anthropic. One of the sector's leading companies has therefore already taken a similar step, and OpenAI is now joining it.
The "textGrain" system
For the labelling, OpenAI uses its own proprietary system called "textGrain". The available excerpt of the source indicates that it affects the model's choice of words, but technical details are not included in this part of the material.
The available source text gives no further information on implementation timelines or the exact way the system works.
