OpenAI introduces text watermarking for EU users to meet regulatory requirements

OpenAI will begin adding an invisible, machine-readable watermark to text generated by ChatGPT and Codex in the European Union, TechCrunch AI and The Verge AI reported. The company stated in a blog post on Monday that the change is intended to comply with the transparency rules of the EU AI Act, which took effect on August 2. TechCrunch AI and The Verge AI both reported that the watermark, called textGrain, will roll out over coming weeks to eligible ChatGPT and Codex users on all plans in the EU. [1][3][5][8]

For developers using OpenAI's API worldwide, TechCrunch AI reported that the feature can be turned on for select models starting today, though it remains off by default. The Verge AI added that OpenAI said textGrain will not be the global default for now. OpenAI noted that editing can make the invisible marks harder to detect, according to TechCrunch AI. The Verge AI reported that OpenAI claims textGrain matched or exceeded rival watermarking systems, such as Google DeepMind's SynthID. [1][3]

ChatGPT is a generative artificial intelligence chatbot originally released by OpenAI in November 2022. The system uses generative models to produce text, code, images, and other digital data based on user prompts. The EU AI Act requires artificial intelligence companies to mark generated content so that other systems can identify it. [1][10][11]