OpenAI is adding text watermarking in ChatGPT and Codex
OpenAI watermarks ChatGPT output for EU compliance—but the feature won't stop determined scraping, and adoption beyond Europe remains unclear.
Why it matters
OpenAI is rolling out invisible, machine-readable watermarks (textGrain) to ChatGPT and Codex text output in the EU to comply with the AI Act. The watermark is claimed to match or exceed competing approaches like Google DeepMind's SynthID and Anthropic's August announcement, with benchmarks showing minimal performance impact—but the feature does not guarantee provenance or prevent misuse, and it remains EU-only for now.
The key facts
8 to knowtextGrain watermark now rolling out to ChatGPT and Codex, EU-first
Invisible, machine-readable watermark designed for AI Act compliance
OpenAI claims textGrain matched or exceeded SynthID and Anthropic's watermarking approaches
Benchmarks show watermarked and unwatermarked text performed similarly
OpenAI explicitly states watermark 'does not guarantee' provenance or prevent misuse
Same compliance driver as Anthropic's August announcement
Scope: EU region, no timeline announced for broader rollout
Not everyone reacted positively to the feature addition
The story so far
Earlier coverage of this storyline
Go to the source
The Verge AItheverge.com
Publisher excerpt: An invisible, machine-readable watermark in text output is rolling out to ChatGPT and Codex, but only for users in the European Union at first. OpenAI says its textGrain watermarking "matched or exceeded" other approaches like Google DeepMind's SynthID for text, which is also the basis for the…