OpenAI is adding text watermarking in ChatGPT and Codex
OpenAI is rolling out textGrain watermarking for ChatGPT and Codex in the EU to meet AI Act rules.
OpenAI is introducing textGrain, an invisible machine-readable watermark, for ChatGPT and Codex users in the European Union over the coming weeks to meet EU AI Act and Code of Practice requirements. It will not be a global default at launch. API customers worldwide can opt in for select models, and approved researchers can request a detector that only reports whether an OpenAI watermark is present. OpenAI says textGrain matched or exceeded Google DeepMind's SynthID and Anthropic's watermark, with similar benchmark scores, but detection is not guaranteed and does not prove authorship.
- EU ChatGPT and Codex users get textGrain over coming weeks.
- Watermarking will not be a global default at launch.
- Worldwide API customers can opt in for select models.
- Detector access is limited to approved researchers, not the public.
- OpenAI says detection is imperfect and does not prove authorship.
Full article374 words · extracted from theverge.com · click to collapse
Stevie Bonifield
is a news writer covering all things consumer tech. Stevie started out at Laptop Mag writing news and reviews on hardware, gaming, and AI.
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 watermarking Anthropic announced in August. Like OpenAI, Anthropic made the move to meet the requirements of the EU’s AI Act; however, not everyone was happy to learn about the addition.
OpenAI also included scores from AI benchmarks showing similar performance from watermarked and unwatermarked text. But it notes that textGrain “does not guarantee reliable detection,” and says text watermarks don’t verify accuracy, determine who owns the text, measure how much a human contributed, or prove human authorship.
Rolling out text watermarking in the EU. Over the coming weeks, we will introduce text watermarking to eligible ChatGPT and Codex users across all plans in the EU only. We are not making text watermarking a global default at launch. This regional approach gives us room to learn from real-world use and feedback.
Enabling opt-in watermarking for API customers. Starting today, API customers around the world will be able to opt in to watermarked text outputs for select models. This lets customers decide how watermarking fits their transparency obligations and the experiences they provide to users. We are also working with cloud partners to make watermarking available for OpenAI model outputs accessed through their services in the coming weeks.
Providing detector access to researchers and expert organizations. Approved researchers and expert organizations can apply starting today. In accordance with the Code of Practice, access will be initially granted on a case-by-case basis to support evaluation and improvement of text provenance. The tool will report whether it detects an OpenAI watermark, without identifying the user or revealing their prompts or conversations. Given the risk of missed watermarks and false positives, we are not making it publicly available at launch.
Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.
- Stevie Bonifield