OpenAI has introduced a large-scale model text watermarking technology called textGrain. Based on the optimal transport theory from over 200 years ago, this technology can embed key information known only to the authorities into AI-generated text without affecting readability or significantly compromising model performance. For texts with a length of 400 tokens, the detection rate exceeds 90% when the false positive rate is 1%. This watermarking technology is immediately available to API customers worldwide and will subsequently be integrated into products such as ChatGPT. Currently, the detector is only accessible to approved researchers and professional institutions, while the technology itself will be open-source. The technology was developed by a team spanning fields such as statistics and economics, with core members primarily from the School of Mathematical Sciences at Peking University. With Google, Anthropic, and OpenAI all implementing AI text attribution watermarking, the era of AI-generated writing with real-name system (real-name registration) has officially arrived. In the future, if users directly utilize AI-generated content, it will be traceable, creating a new demand for proving content as personally original.