Google Messages, the popular messaging app, is in the process of developing a tool that can identify whether an image shared in a conversation has been created or modified by artificial intelligence (AI). Evidence of this feature was uncovered in the app’s source code, indicating that it will take advantage of C2PA content credentials – integral data that can verify the creation and manipulation history of an image file. Regardless of the device used, this feature is expected to be a standard element in the Google Messages application, which is pre-installed on most top-level Android phones.
The discovery of this development was made by AssembleDebug, a skilled code investigator who has a history of revealing Google’s yet-to-be-released features. The initial groundwork for this tool has been detected in various recent releases of the app, although it isn’t operational at this time. It should be seen as a significant signal of what’s to come, rather than an available feature.
The Coalition for Content Provenance and Authenticity (C2PA) is the organization behind the standard that records the creation and editing history of an image. Essentially, this would mean your phone could inform you whether a photo you received was taken with a camera or generated via AI.
This aligns with Google’s broader strategy of integrating SynthID watermark detection into Google Photos and Gemini, and now extending this into Google Messages. The tool is expected to have around 18 labels, distinguishing between photos taken with cameras, those modified by partial AI edits, and fully AI-generated media, demonstrating Google’s emphasis on nuanced detection.
This development comes amidst growing apprehension around the increasing sophistication of AI-generated images. Some fear that even those with technical know-how are starting to fall for these hyper-realistic images, underlining the critical need for such detection tools.
However, execution is not without potential pitfalls. Meta Platforms, the company behind Facebook, Instagram, and WhatsApp, implemented a similar feature, which encountered issues due to false positives. For instance, genuine photos were incorrectly flagged as AI-generated due to metadata left behind by standard Photoshop edits. Google Messages’ adoption of the C2PA standard might face similar issues.
The primary beneficiaries of this tool would be users who often receive forwarded images, group-chat memes, and screenshots. A simple tap to verify the image’s provenance would be more efficient than manually scrutinizing the image.
In terms of accessibility, users would be able to reach this feature by clicking on an image, opening the overflow menu, and selecting “View details”. It’s also worth noting that this feature isn’t live yet, so there’s nothing to test currently.
Interestingly, Google, which has developed impressive image-generating models like Nano Banana, is now building a tool to detect the products of such models. Despite the seeming contradiction, it’s a necessary step towards ensuring authenticity and transparency in digital communication. If Google’s tool proves more reliable than Meta’s version, it will be a game-changer.
What is the purpose of the new tool being developed for Google Messages?
The tool aims to identify whether an image shared in a chat has been created or manipulated by AI, helping users distinguish real photos from AI-generated ones.
What is the C2PA standard?
The C2PA (Coalition for Content Provenance and Authenticity) standard records the creation and editing history of an image, allowing verification of whether a photo was captured with a camera or generated by AI.
What potential issues might arise with this tool?
Similar to Meta’s implementation, Google’s tool may encounter issues with false positives, where genuine photos are incorrectly flagged as AI-generated due to the metadata left behind by standard editing tools such as Photoshop.
