Google DeepMind has developed technology to apply invisible watermarks to images generated by artificial intelligence

Google DeepMind has developed SynthID technology to apply invisible watermarks to images generated by artificial intelligence.

Here's What We Know

SynthID embeds an invisible watermark directly onto the pixels of an image. At first, the tool will be available to a limited number of users of Imagen, an image generator created by Google.

According to the developers, the watermark cannot be removed using conventional image editing methods. SynthID is designed not to degrade the quality of the image and to allow the mark to be detected even after applying filters, changing colours and saving with different degrees of compression.

SynthID consists of two deep learning models: one for watermarking and one for identification. The developers said that both models were trained on a variety of data and combined into a single system.

Google admits that the method is not perfect. However, the developers called SynthID a promising approach that will allow generative images to be handled responsibly.

In the future, the company plans to expand the tool to other AI models that generate text, video and audio.

Source: DeepMind