OpenAI will watermark ChatGPT text in the EU but makes it optional for API users worldwide

OpenAI is adding invisible watermarks to ChatGPT text to comply with EU rules, and will enable watermarking for ChatGPT and Codex users in the European Union over the coming weeks. Unlike Anthropic's mandatory Claude watermarking, which applies globally regardless of access method, OpenAI's API watermarking will be opt-in worldwide and will also be available through cloud partners including Microsoft Azure in the coming weeks.

The technology, called textGrain, embeds an invisible statistical signal in the model's word choices, similar to Claude's SynthID watermark, which uses Google's open-source technology. The EU AI Act reportedly requires AI providers to label generated text in a machine-readable format.

OpenAI says textGrain matched or beat other approaches in internal tests, including Google's SynthID for text, and that it plans to release the technology as open source. Detection rates depend heavily on text length and subject matter. With the detector set to a 1 percent target false-positive rate, it identified watermarks in about 95 percent of 400-token psychology passages, falling to about 80 percent at 200 tokens. Detection was substantially lower for math content, where OpenAI says the model has less freedom to choose words, and about 60 percent for math passages at 400 tokens.

Editing reduces detection. Replacing 10 percent of words with synonyms cuts rates for 400-token passages from about 92 percent to 66 percent; replacing a quarter of words drops detection to 17 percent. OpenAI says watermarking does not hurt output quality, citing tests of its frontier model Astra showing no significant performance differences with watermarking on or off across eight benchmarks, including GPQA Diamond, BrowseComp, and DeepSWE. OpenAI also says a detected watermark does not reveal how much human creativity or editing went into text, does not establish ownership, assign responsibility, identify a user, or show accuracy, and that failing to detect one does not prove human authorship.

Detector access will initially be limited to selected researchers and specialist organizations, who can apply and be granted access case by case under the EU's Code of Practice. The tool reports only whether it detected an OpenAI watermark, not users, prompts, or conversations. OpenAI says it is restricting access because the detector can flag unmarked text or miss watermarks, and plans to expand access when results can be interpreted responsibly, without saying when. Verification tools for images and audio remain publicly available, including openai.com/verify and the Content Provenance API.