Please enable JavaScript to experience the full functionality of GMX.

Google tests digital watermark to spot AI created images

Google tests digital watermark to spot AI created images

Google is testing a digital watermark to spot images created by artificial intelligence (AI).

The tech giant's AI arm DeepMind has developed SynthID to identify images generated by machines as they look to battle disinformation.

The programme embeds changes to individual pixels in images which means watermarks can be detected by computers but remain invisible to the human eye.

However, DeepMind warned it's not "foolproof against extreme image manipulation".

Tools such as Midjourney - which has over 14.5 million users - let people create images in a matter of seconds through simple text instructions.

Google has its own image generator, Imagen, and the system for creating and checking watermarks will just apply to images created with this tool.

The use of AI image generators have raised issues over copyright and ownership around the world.

Meanwhile Pushmeet Kohli, head of research at DeepMind, told the BBC the system makes such a simple modification to images "that to you and me, to a human, it does not change".

He explained that the watermark will still be identifiable, no matter how much the image is cropped or edited.

He added: "You can change the colour, you can change the contrast, you can even resize it... [and DeepMind] will still be able to see that it is AI-generated."

Sponsored Content

Related Headlines