<figure>
<img alt="Open AI logo on a graphic green background." src="https://platform.theverge.com/wp-content/uploads/sites/2/2025/10/STK155_OPEN_AI_CVirginia__C.jpg?quality=90&strip=all&crop=0,0,100,100" /> <figcaption> </figcaption> </figure> <p>An <a href="https://openai.com/index/eu-text-provenance/">invisible, machine-readable watermark</a> in text output is being rolled out to ChatGPT and Codex, initially for users in the European Union.</p> <p>OpenAI reports that its textGrain watermarking technology "matched or exceeded" other techniques, such as Google DeepMind's SynthID for text, which was also the basis for watermarking announced by Anthropic in August. Both OpenAI and Anthropic implemented these measures to comply with the <a href="https://www.theverge.com/ai-artificial-intelligence/974571/eu-ai-act-transparency-labels-rules-deepfakes">EU's AI Act</a>. However, there has been some dissatisfaction regarding this addition.</p> <p>OpenAI has also provided scores from AI benchmarks indicating similar performance between watermarked and unwatermarked text. It is important to note that textGrain "does not guarantee ...</p> <p><a href="https://www.theverge.com/ai-artificial-intelligence/1004880/openai-chatgpt-text-watermarks-eu-ai-act">Read the full story at The Verge.</a></p>