OpenAI adds a text blind watermark to the EU ChatGPT and Codex. The relevant detectors are only open to professional institutions.

#AI Based on legal requirements, OpenAI has enabled a text blind watermark mechanism for ChatGPT and Codex in the EU market, which can identify whether the content is generated by AI. However, OpenAI does not plan to disclose the detector to the public for the time being, and only qualified institutions can apply for detector permission. The blind watermark mechanism does not use certain invisible characters, but adjusts the probability of the model selecting different words to form a signal, so only rewriting the content fragments will reduce the recognition rate. View the full text: https://ourl.co/127177