Google Adds Digital Watermarks to AI-Edited Images

Google Introduces Watermarks for AI-Edited Photos

Google announced on Thursday the implementation of a digital watermark system for images within Google Photos. This change, taking effect this week, will apply to photos modified using generative AI.

The watermark will be specifically visible on images altered through the Reimagine feature, which is part of the Magic Editor available on Pixel 9 devices.

Reimagine and the Rise of AI Image Editing

Introduced in 2024, Reimagine leverages generative AI to modify various elements within a digital photograph. It’s intended primarily for refining images captured directly on the device.

However, the capabilities of Reimagine extend to more substantial image manipulations, potentially obscuring the distinction between authentic photos and those entirely created by AI.

The Need for Transparency in AI-Generated Content

With the increasing realism of AI-generated images, there's a growing demand for a standardized method to help the public identify content created artificially. Digital watermarking is frequently proposed as a viable solution.

This technique allows for the addition of identifying information to a file without directly impacting the visual appearance of the image itself.

SynthID: Google’s Watermarking Technology

Google will utilize SynthID to mark images created or altered with AI. Developed by Google’s DeepMind, SynthID embeds a digital watermark directly into the content.

Importantly, this embedding process doesn't compromise the integrity of the original image. SynthID also provides a tool for scanning images to detect the presence of these watermarks.

Currently in beta, SynthID is also compatible with AI-generated text and video files.

Accessing Watermark Information

Users can access information about the digital watermark by selecting “About this Image” within Google Photos. This will reveal the watermark embedded in the image’s metadata.

It’s important to note that not all edits will trigger the application of SynthID. Minor alterations, such as changing the color of a small object in the background, may be too subtle to be detected and labeled.

A Commitment to AI Transparency

This deployment aligns with Google’s broader initiative to enhance transparency surrounding AI editing within Google Photos. The company’s AI principles serve as the foundation for these decisions.

Google aims to provide users with clear indications when images have been modified using AI technology.