Schimbare importantă pentru ChatGPT în Europa: OpenAI introduce watermark în textele generate de AI

OpenAI plans to introduce an invisible watermark in texts generated by ChatGPT and Codex for EU users, aiming to comply with European regulations. The technology, named 'textGrain', embeds a statistical signal in the text without visible markers. A special detector can later identify this signal to determine if a text contains the watermark. The implementation will begin on October 5, with options for API clients worldwide, while EU users will see the watermark in the coming weeks. However, OpenAI warns that the system has limitations, including the potential for false positives and the inability to definitively establish text origin or ownership. The watermark does not guarantee information accuracy and will initially be available only to approved researchers and organizations.