Google launches watermarks for AI-generated images
Spread the love

The watermark is embedded directly into images created by Imagen, one of Google’s latest text-to-image generators, using a technology called SynthID. No matter how many filters or colors are added, the AI-generated label remains the same.

With SynthID, you can also scan incoming images to determine whether they were made by Imagen by scanning for the watermark with three levels of certainty: detected, not detected, and possibly detected.

In a blog post Tuesday, Google wrote, “While this technology isn’t perfect, it is accurate against many common image manipulations.”

Vertex AI, Google’s generative-AI platform for developers, now offers a beta version of SynthID. SynthID, created by Google’s DeepMind unit and Google Cloud, will continue to evolve and may expand to other Google products. Digital watermark efforts have been led by the Coalition for Content Provenance and Authenticity (C2PA), an Adobe-backed consortium, while those of Google have been largely independent.

Earlier this year, Google introduced a tool that allows users to see when images found on its site were originally indexed by Google, where they first appeared, and where else they can be found.

Additionally, Google announced that its AI-generated images will carry a markup in the original file to “give context” if they appear on another website or platform.

It’s unclear whether these technical solutions can fully address the problem as AI technology develops faster than humans can keep up with it. Earlier this year, OpenAI, the company behind Dall-E and ChatGPT, admitted that its own effort to detect AI-generated writing is “imperfect” and should be “taken lightly.”