Google releases SynthID watermark detector for AI-generated media

Source: Google. — Image credit: Search Engine Journal

MOUNTAIN VIEW, Calif. — Google made its SynthID watermark detector, designed to identify AI-generated media, publicly available in English Tuesday.

The tool allows users globally to upload image, video and audio files to synthid.com to check for invisible watermarks embedded by Google’s AI models and those of its partners, according to the company.

SynthID works by detecting a digital watermark that is imperceptible to the human eye or ear but can be identified by the detector, Google DeepMind said.

The system is specifically designed for media created using AI from companies that have adopted SynthID technology, including OpenAI, NVIDIA and Kakao, with Apple expected to join soon.

Google emphasized that SynthID is not a general AI detector and will only recognize watermarks from its own AI or that of its partners.

Detection results indicate whether media was “made or edited with Google AI” and may also note if it has been “edited further since then.”

Pushmeet Kohli, a research scientist at Google DeepMind, previously stated that the technology aims to provide a reliable method for distinguishing AI-generated content from human-created media.

For optimal accuracy, the company advises users to upload the best-quality file available and to gather multiple data points to make decisions about media authenticity.

The public release follows an earlier limited rollout, marking a significant step in Google’s efforts to provide tools for identifying the provenance of AI-generated content.

Source: Search Engine Journal