Sunday, November 24, 2024

Google Plans To Expose AI-Generated Images In New Google Photos Update

Must read

Google is rolling out a significant new update that will make it easier to spot pictures that were generated or edited with AI.

Starting this week, Google Photos will add a new “AI info” section to the image details display letting users know whether the image they’re viewing has been created or manipulated using AI. By checking this information, users will be able to see more easily when an image has been generated with AI tools, such as Google’s Gemini app, or manipulated with AI-powered functions like Google Photos’ Magic Eraser.

First leaked earlier this month by tipster Assemble Debug, these new AI transparency features have now been officially announced in a recent Google Photos blog post that details how the company will make it easier for users to spot AI in images.

As revealed in the screenshot above from Android Authority, the new AI info section includes “Credit: Made with Google AI” and “Digital source type: Created using Generative AI,” revealing that the selected image is an AI generation rather than a real photo. These details come from information known as IPTC metadata, which is optionally embedded in the image files whenever they are saved or edited.

According to Google, you’ll also see when tools such as Magic Editor, Magic Eraser and Zoom Enhance have been used. And, because IPTC metadata is an industry-standard format, you’ll be able to see similar AI information disclosed by apps from many other companies, including Adobe and Microsoft.

Why Is AI Info Important?

Amid increasing concerns over the potential misuse of AI, Google’s latest steps will increase transparency and awareness of the tools available and give users a first line of defense against image-based misinformation and deepfakes. If you’re not sure about an image found online, you can download it and open it in Google Photos to see quickly whether AI tools have been used before making a decision about its authenticity.

The change may also cause users to think twice before sharing manipulated images, as their recipients may be alerted to the fact that a picture has been edited. This information could have positive effects on stamping out misinformation and deepfakes or projecting a realistic body image online.

A Good Start, But Not A Solution To The Problem

While Google’s latest move will undoubtedly bring greater transparency to users, it’s important to note that Google is simply passing along information already stored voluntarily within the images. While Google’s AI-powered tools embed this information automatically, other tools may not. Furthermore, anyone determined to conceal their use of AI will remain able to do so with ease as it’s trivially easy to edit or remove IPTC metadata before sharing an image—simply taking a screenshot of the picture would do. Many online services automatically strip out such data when content is uploaded, so it’s not only possible but likely that this “AI Info” will become lost on its digital journey from the original creator to you. Conversely, it’s just as easy for a bad actor to add fake AI tags to a real photo to discredit it.

However, more sophisticated techniques are under development. Google is already pioneering more robust technologies, such as SynthID, which embeds invisible watermarks into AI-generated images, video, text or audio, which can be decoded later using software tools.

These watermarks are considerably more challenging to remove as they are merged deeply with the actual media they protect rather than as a simple tag that could be removed or changed. For example, removing an artist’s name from a label next to a painting would be easy, but concealing the creator’s style would be much more complicated as it’s an intrinsic part of the work. SynthID’s watermarks are like imperceptible tell-tale brushstrokes, detectable only by a skilled forensic art analyst.

You will find Google Photos’ new AI Info by selecting a photo and swiping up in the app or by tapping the (i) icon displayed in the web version at photos.google.com.

Follow @paul_monckton on Instagram.

ForbesGoogle Photos To Help Users Identify AI-Created ImagesForbesGoogle Enforces Essential New Android Photo Privacy Feature

Latest article