Google Launches Tech to Watermark AI Generated Text

Google has announced that its SynthID Text technology, which enables developers to watermark and detect text generated by AI models, is now generally available. This tool can be accessed via the AI platform Hugging Face and Google’s updated Responsible GenAI Toolkit.

In a recent post on X, Google stated, “We’re open-sourcing our SynthID Text watermarking tool. Available freely to developers and businesses, it will help them identify their AI-generated content.”

How Does SynthID Text Work?

SynthID Text utilizes a unique method to watermark generated text. When a prompt is given—such as “What’s your favorite fruit?”—text-generating models predict the next token (a character or word) one at a time. Each token is assigned a probability score indicating its likelihood of being included in the output. SynthID Text modifies this token distribution by adjusting the likelihood of specific tokens being generated.

According to Google, “The final pattern of scores for both the model’s word choices combined with the adjusted probability scores are considered the watermark.” This pattern is then compared to expected patterns for both watermarked and unwatermarked text, allowing SynthID to detect whether the text was produced by an AI tool or sourced from elsewhere.

Google asserts that SynthID Text, integrated with its Gemini models since spring, does not compromise the quality, accuracy, or speed of text generation. It remains effective even on text that has been cropped, paraphrased, or modified.
Importance of Watermarking Technology

As misinformation and inappropriate content generated by AI become increasingly prevalent, watermarking technology is gaining importance. Google’s SynthID aims to assist developers in identifying AI-generated content responsibly. This initiative comes at a time when regulatory discussions around mandatory AI watermarks are emerging in various regions.

Overall, Google’s open-source SynthID Text tool represents a significant step toward enabling developers to better manage AI-generated content while maintaining high standards of quality and integrity.

Watermark gif

Video reference