ChatGPT will soon start secretly watermarking its answers — here's what that means for the text you copy and share (www.techradar.com)

OpenAI has announced plans to implement invisible watermarking for ChatGPT text outputs in the European Union, aligning with the EU AI Act's requirements for AI-generated content to be identifiable. This watermarking process, developed through OpenAI's textGrain technology, subtly alters word choices to embed a statistical signal in the text, which can be detected by special AI tools provided to select researchers. However, users outside the EU, including those in the US and UK, will not be affected by this change. This move is significant as it highlights the compliance challenges AI companies face with regulatory frameworks and raises concerns within the AI community about altering text generation for the sake of watermarking. Critics argue that such a system may degrade the quality of generated content and could be easily circumvented. The limitations of detection tools, which may produce false positives or fail with shorter texts, further complicate the effectiveness of this watermarking strategy. Overall, this initiative appears to be more of a regulatory checkbox than a robust solution, prompting discussions about the balance between transparency and the quality of AI-generated responses.