Independent AI research publications and media reports

Google Expands SynthID Watermark to AI-Generated Video and Text, Announces DeepMind CEO Demis Hassabis

May 14, 2024, 06:44 PM

Google has announced the expansion of its SynthID watermark imprinting system to include AI-generated video and text. Previously, SynthID was used to add imperceptible watermarks to AI-generated images and audio, making them easier to distinguish. The upgraded system, announced by DeepMind CEO Demis Hassabis, aims to help identify generative text and video. Google will also open-source the SynthID text watermarking through its updated Responsible Generative AI Toolkit, enabling more developers to build AI responsibly. The new expansion includes the Veo model, which will also incorporate these imperceptible watermarks.