OpenAI is introducing a system designed to help identify AI-generated text while also expanding advertising tests on ChatGPT. The text watermarking feature will initially cover the European Union, while a new ad format will be tested in the US later this month.
OpenAI to watermark AI-generated text
Over the coming weeks, text generated by ChatGPT and Codex in the EU will carry an invisible statistical watermark. API customers worldwide can also enable the feature for select models, although text watermarking will remain disabled by default.
OpenAI said the move responds to the EU AI Act, which requires providers of generative AI systems to make machine-generated content identifiable. The company has also started accepting applications from researchers and expert organisations to test its text watermark detector.
The system, called textGrain, adds an invisible statistical pattern through the model’s word choices. A detector searches for the signal to assess whether text contains an OpenAI watermark. OpenAI said textGrain matched or exceeded other watermarking methods tested, including Google’s SynthID for text. It also plans to release the underlying technology as open source.
OpenAI acknowledged that editing AI-generated text can reduce detection accuracy. In its evaluation, the detector identified around 80% of watermarked 200-token passages at a 1% target false-positive rate. Detection rose to roughly 95% for 400-token passages in areas such as psychology, while results were less consistent for mathematics.
For 400-token passages, replacing 10% of words with synonyms reduced detection from around 92% to 66%. Replacing 25% of words reduced it to 17%.
The company said the watermark cannot measure how much human input was used or determine ownership. It can indicate OpenAI involvement but should not be treated as proof of authorship.
ChatGPT to test new ad format
OpenAI will also test a “new visual ad format” later this month, initially during image generation in ChatGPT. The ads will be labelled and kept separate from generated images.
The ads may show details about products, how they are used or experiences they enable, creating new opportunities for advertisers. The initial test will take place in the US with a group of advertisers.
“As with our other ad formats, advertising does not influence the answers ChatGPT provides”, OpenAI said.
Also read: Viksit Workforce for a Viksit Bharat
Do Follow: The Mainstream LinkedIn | The Mainstream Facebook | The Mainstream Youtube | The Mainstream Twitter
About us:
The Mainstream is a premier platform delivering the latest updates and informed perspectives across the technology business and cyber landscape. Built on research-driven, thought leadership and original intellectual property, The Mainstream also curates summits & conferences that convene decision makers to explore how technology reshapes industries and leadership. With a growing presence in India and globally across the Middle East, Africa, ASEAN, the USA, the UK and Australia, The Mainstream carries a vision to bring the latest happenings and insights to 8.2 billion people and to place technology at the centre of conversation for leaders navigating the future.