OpenAI to Watermark ChatGPT Text in the European Union
OpenAI is introducing invisible watermarking for text generated by ChatGPT and Codex within the European Union, aligning with regulatory requirements under the EU AI Act.
Radwan AbdulhadiEditor and technology writer
Compliance with the EU AI Act
OpenAI has announced that it will begin implementing an invisible watermark for text and code generated by ChatGPT and Codex in the European Union. According to an official company <a href="https://openai.com/index/eu-text-provenance/">blog post</a>, this rollout is designed to comply with transparency regulations established under the legislation.
The mandate falls under <a href="https://digital-strategy.ec.europa.eu/en/policies/guidelines-ai-transparency-obligations">The EU AI Act’s transparency rules</a>, which require artificial intelligence developers to ensure that AI-generated content can be readily identified by other digital systems.
How the textGrain Watermarking Method Works
The watermarking approach, detailed in a newly published <a href="https://cdn.openai.com/pdf/e9508624-d767-41b6-a26d-e34ca798ada6/textgrain-entropy-calibrated-watermarking-for-language-model-text.pdf">technical report</a> co-authored with researchers from Yale and the University of Pennsylvania, relies on a proprietary system named textGrain.
Rather than displaying a visible symbol or badge, the technology subtly shapes a model's word choices during generation. This leaves an invisible statistical pattern that remains embedded within the text, traveling along even when passages are copied and pasted across different applications.
OpenAI noted that the system does not track or identify individual users, and internal testing showed no meaningful impact on model performance when the feature is enabled.
Rollout Strategy and API Availability
The feature is scheduled to roll out over the coming weeks to eligible ChatGPT and Codex users across all account tiers, though its activation is limited strictly to the European Union region.
For developers utilizing <a href="https://openai.com/index/eu-text-provenance/">OpenAI</a>'s API globally, the watermark can be manually enabled for select models. However, the company opted not to enable text watermarking as a global default setting during this initial launch phase.
Detection software access will initially be tightly controlled. OpenAI stated that it plans to grant detector access primarily to approved researchers and expert organizations to evaluate reliability and appropriate use cases.
Limitations and Editing Challenges
Internal tests conducted by the company revealed certain technical limitations regarding the robustness of the watermarks. For instance, replacing roughly 10% of generated words with alternative synonyms caused the detection rate to drop significantly from approximately 92% down to 66%.
Furthermore, short passages, translated text, and mathematical answers were found to be notably more difficult to detect accurately. Because of these variables, OpenAI cautioned that a missing watermark does not definitively prove human authorship, as text can be heavily edited, shortened, or generated by alternate AI systems.
Industry Context and Prior Developments
The decision follows a <a href="https://techcrunch.com/2026/08/11/anthropic-says-it-will-watermark-text-generated-by-its-ai-models/">Anthropic said it would watermark</a> announcement made a couple of months prior regarding its Claude models, which sparked widespread conversation across the sector.
That prior rollout <a href="https://techcrunch.com/2026/08/12/some-claude-users-are-mad-that-anthropics-new-watermarks-will-catch-them-cheating-at-their-jobs-classes/">drew backlash from some Claude users</a> who felt that users provided the core context and direction while the software simply served as an execution tool.
Prior reporting by <a href="https://www.wsj.com/tech/ai/openai-tool-chatgpt-cheating-writing-135b755a">The Wall Street Journal reported</a> in 2024 that OpenAI had previously developed text watermarking technology but delayed release due to competitive concerns that users might migrate to platforms without such restrictions.
Major industry stakeholders, including Anthropic, Google, Meta, Microsoft, and OpenAI, have all committed to complying with the broader <a href="https://digital-strategy.ec.europa.eu/en/news/strong-backing-code-practice-transparency-ai-generated-content">EU’s code of practice</a> governing artificial intelligence-generated content.
Sources
- TechCrunchOpenAI will start watermarking ChatGPT’s text in the EU
- EngadgetOpenAI will add a digital watermark to text and code generated in the EU
Continue chronologically
Related entity coverage
Related stories
OpenAI Introduces Text Watermarking in ChatGPT and Codex
OpenAI is rolling out invisible machine-readable watermarks for ChatGPT and Codex in the EU to comply with transparency rules.
OpenAI Details Text Watermarking Approach for EU AI Act
OpenAI announces its phased text watermarking approach under the EU AI Act, featuring textGrain technology, API opt-ins, and researcher access.
OpenAI Hosts DevDay 2026 With Over 20 Major Announcements
OpenAI highlights more than 20 major announcements at DevDay 2026, introducing new agents, Codex updates, models, and builder tools.
OpenAI Expands Academy with Role-Specific AI Learning Paths
OpenAI has expanded its OpenAI Academy to include specialized learning paths for developers, leaders, educators, and students to build practical AI skills.
