.png)
April 27, 2025
OpenAI, known for its AI chatbot ChatGPT, is experimenting with implementing watermarks in AI-generated texts. Similar to images created with the AI model DALL-E, these markings are intended to enable the identification of machine-written content. This step raises questions about transparency and the impact on the use of AI-generated texts.
The watermarks consist of invisible Unicode characters embedded in the text. Specifically, it's the "Narrow No-Break Space" (Unicode U+202F) character, which visually resembles a normal space. These characters are copied along with the text and are thus retained, even if the text is pasted into other applications or platforms. By analyzing the character string, it can theoretically be determined whether a text was generated by ChatGPT.
However, the effectiveness of this method is controversial. Experts point out that the watermarks are relatively easy to remove. Using common word processing programs and the find-and-replace function, the special Unicode characters can be replaced with normal spaces. Thus, the marking is only effective as long as the method of implementation and removal remains unknown to most users.
OpenAI argues that the watermarks, despite their imperfection, make an important contribution to the responsible use of AI. The marking is intended to help curb misuse such as the spread of misinformation or deception about the authorship of texts. Critics, however, fear that the watermarks could restrict the free use of AI-generated texts and lead to surveillance of users. The discussion about the balance between transparency and data protection in the context of AI-generated content continues.
The development of watermarks for AI-generated texts is still in its early stages. It is expected that OpenAI and other companies will work on more robust methods that are more difficult to circumvent. At the same time, tools are being developed to facilitate the detection of AI-generated texts. The implications of these developments for the future of communication and copyright are far-reaching and require careful consideration.
Mindverse, the German provider of AI-based content solutions, is closely following the developments in the field of AI-generated texts. As an expert in customized AI solutions, including chatbots, voicebots, AI search engines, and knowledge systems, Mindverse offers companies the opportunity to use the potential of AI effectively and responsibly. With a holistic approach that encompasses text, images, research, and more, Mindverse supports companies in mastering the challenges of digital transformation.
Sources: - t3n.de: Invisible characters in ChatGPT texts – why OpenAI relies on hidden watermarks - Facebook: t3n Magazine Post about watermarks in ChatGPT texts - fr.de: ChatGPT texts have secret watermarks - finanznachrichten.de: OpenAI hides invisible characters in ChatGPT texts - futurezone.at: How to remove watermarks from ChatGPT texts - LinkedIn: t3n Magazine Post about watermarks in ChatGPT texts - t3n.de: Artificial Intelligence - chip.de: Attention, ChatGPT incorporates invisible traces into the texts