OpenAI is introducing an invisible, machine-readable text watermark called textGrain in outputs from ChatGPT and Codex, but the feature will initially be limited to users in the European Union. The company said the rollout applies to eligible ChatGPT and Codex users across all plans in that region over the coming weeks, and that it is not making text watermarking a global default at launch. OpenAI said the regional approach gives it room to learn from real-world use and feedback.
OpenAI said textGrain matched or exceeded other approaches, including Google DeepMind's SynthID for text, which also serves as 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. OpenAI also included scores from AI benchmarks showing similar performance from watermarked and unwatermarked text. The company said textGrain does not guarantee reliable detection.
OpenAI said text watermarks do not verify accuracy, determine who owns the text, measure how much a human contributed, or prove human authorship. The company said it is 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, letting customers decide how watermarking fits their transparency obligations and the experiences they provide to users. OpenAI said it is also working with cloud partners to make watermarking available for OpenAI model outputs accessed through their services in the coming weeks.
OpenAI said it is providing detector access to researchers and expert organizations. Approved researchers and expert organizations can apply starting today, and access will initially be granted on a case-by-case basis in accordance with the Code of Practice 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, OpenAI said it is not making the tool publicly available at launch.
The addition follows the EU's AI Act, which drove Anthropic's August watermarking announcement and OpenAI's move. OpenAI's earlier textGrain announcement drew mixed reactions, with not everyone happy to learn about the addition. The US market is not part of the initial ChatGPT and Codex rollout, though API customers worldwide can opt in, and cloud partner availability is planned in the coming weeks.
More AI news from TechManNews.
