OpenAI Rolls Out textGrain Watermarking in EU, Detector for Researchers
OpenAI is bringing text watermarking to ChatGPT and Codex in the European Union while offering API customers an opt-in option worldwide.
OpenAI announced it is rolling out its textGrain text watermarking system to eligible ChatGPT and Codex users on all plans in the European Union over the coming weeks, according to The Verge. The company is not making text watermarking a global default at launch.
Starting today, API customers worldwide can opt in to receive watermarked text outputs for select models. OpenAI also said it is working with cloud partners to make watermarking available for OpenAI model outputs accessed through their services in the coming weeks.
Approved researchers and expert organizations can apply starting today for access to a watermark detection tool, which will be granted on a case-by-case basis initially. The tool reports whether an OpenAI watermark is detected without identifying the user or revealing prompts or conversations.
OpenAI said it is not making the detector publicly available at launch due to risks of missed watermarks and false positives. The company stated that its textGrain watermarking matched or exceeded other approaches, including Google DeepMind's SynthID for text.
Background
In August, Anthropic announced watermarking with SynthID as its basis to meet EU AI Act requirements.
Quick answers
Which users will get textGrain watermarking?
Eligible ChatGPT and Codex users on all plans in the European Union will receive it over the coming weeks. API customers worldwide can opt in for select models starting today.
Can the public use the watermark detector?
No. OpenAI is not making the detector publicly available at launch. Approved researchers and expert organizations can apply starting today, granted case-by-case initially.
How does OpenAI's textGrain compare to SynthID?
OpenAI says its textGrain watermarking matched or exceeded other approaches like Google DeepMind's SynthID for text.