OpenAI will watermark ChatGPT text in the EU but makes it optional for API users worldwide
OpenAI watermarks ChatGPT in the EU—but API customers can opt out. Anthropic doesn't give that choice.
Why it matters
OpenAI is deploying invisible textGrain watermarks on EU ChatGPT output to comply with emerging AI provenance rules, but unlike Anthropic, it allows API customers worldwide to disable them. Detection robustness drops sharply when text is edited, creating a compliance vs. usability tradeoff that affects enterprise deployments.
The key facts
11 to knowInvisible textGrain watermarks added to ChatGPT in EU only
API customers worldwide can opt out of watermarking (Anthropic does not offer this)
Detection rates tested at 95% under clean conditions
Detection drops to 17% when 25% of words are replaced
Watermark survives partial editing but degrades significantly with paraphrasing
Watermarking: invisible Grain watermarks added to ChatGPT in EU
Opt-out available globally for API customers (unlike Anthropic's mandatory approach)
Detection rate: 95% baseline; degrades to 17% when 25% of words are replaced
Regulatory driver: EU compliance requirement (implicit)
No pricing or quota impact disclosed
No rollout timeline specified
Go to the source
The Decoderthe-decoder.com
Publisher excerpt: OpenAI is adding invisible textGrain watermarks to ChatGPT in the EU, but unlike Anthropic, it will let API customers worldwide opt out. Tests show detection rates as high as 95 percent, dropping to 17 percent when a quarter of the words are replaced.