Breaking

OpenAI to Add textGrain Watermarks to ChatGPT and Codex Text in the EU

October 5, 2026 at 13:44 EDT

Policy & Regulation
Security
Foundation Models

OpenAI will add invisible text watermarks to eligible ChatGPT and Codex output in the European Union over the coming weeks, while opening opt-in watermarking for API customers worldwide. The move introduces OpenAI's textGrain system as the company responds to EU rules requiring AI-generated text to be machine-readable and detectable. OpenAI European Commission

October 5, 2026 | OpenAI

Text provenance gets a regional rollout

OpenAI is adding invisible text watermarks to eligible ChatGPT and Codex output in the EU, while global API customers can opt in. The company is keeping detector access limited because results can be missed or misread.

False-positive target

~80%

Detection at 200 tokens

~95%

Detection at 400 tokens

17%

Detection after 25% synonym edits

Editing erodes the signal

In 400-token passages, synonym replacement sharply reduced detection.

Rollout path

EU apps

Eligible ChatGPT and Codex text

API choice

Global opt-in for select models

Detector gate

Approved researchers and experts first

What it helps with

Machine-readable provenance for eligible text
Customer-level choice for API outputs
No user, prompt or account encoded

What remains fragile

Short text and code are harder
Editing or translation can weaken detection
Results do not prove authorship or accuracy

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.