OpenAI Launches Text Watermark Feature for ChatGPT and Codex
Text watermarking technology is being gradually rolled out for ChatGPT and Codex to comply with European Union regulations.
OpenAI has implemented text watermarking for ChatGPT and Codex users to meet the requirements introduced by the European Union Artificial Intelligence Act.
Step Towards Compliance with European Union Regulations
OpenAI has begun a limited deployment of text watermarking and detection technology in the region to fulfill the requirements of the European Union Artificial Intelligence Act.
Over the coming weeks, this invisible and machine-readable watermark called textGrain will be made available to eligible users across all plans.
Options for Global API Users
API customers worldwide now have access to a system where they can optionally choose watermarked text outputs on certain models.
This watermarking feature will remain disabled by default on the API side and will not be a mandatory global standard.
Detector Access for Researchers
The company has officially opened the application process for approved researchers and expert organizations to apply for the text watermark detection tool.
These designated accesses will be evaluated on a case-by-case basis in order to conduct security and oversight processes transparently.
Limitations of the Technology and Other Initiatives
Stating that artificial intelligence watermarking technologies have significant limitations, the company emphasized that this method does not guarantee reliable detection.
Similarly, other artificial intelligence companies such as Anthropic have announced comparable measures to comply with the new European Union regulations.