Google LLC plans to implement C2PA, a technology for verifying the authenticity of media files, in its search engine and advertising systems.
The company detailed the initiative today.
The C2PA, or Coalition for Content Provenance and Authenticity, is an industry group that was formed in 2021 by Intel Corp., Apple Inc. and a number of other companies. Google joined the consortium earlier this year. C2PA develops a technical standard of the same name that makes it possible to determine whether an image was generated using artificial intelligence or modified after its creation.
The standard works by attaching a metadata file called a manifest to every image. This manifest specifies whether the image was generated with AI, when it was created and when. It’s also possible to include other information such what edits that were made to a file after its initial release.
Hackers could theoretically modify a tempered image’s C2PA manifest to hide the fact that the image was edited. To prevent that possibility, the technology protects manifests from malicious edits using a cryptographic technique called hashing.
Hashing provides the ability to create a unique digital fingerprint for a file. C2PA leverages the technology to generate a unique fingerprint for images’ manifests. When hackers attempt to modify a photo’s manifest, the manifest’s fingerprint is modified as well, which makes it straightforward to spot tempering attempts.
C2PA includes other cybersecurity features as well. Most notably, its hashing feature generates a unique fingerprint of not only the manifest that is attached to an image but also the image itself. C2PA then places the image’s fingerprint into the manifest. This arrangement binds photos to their respective manifests and makes it impossible for hackers to replace one of the files with a counterfeit or an out-of-date version.
As part of the initiative detailed today, Google plans to implement C2PA in its “About this image” feature. The feature, which is available in its search engine’s Google Images, Lens and Circle to Search tools, provides contextual information about images. When a user selects an image that supports C2PA, the feature will display information on whether the file was created or modified using AI.
Google is also integrating the technical standard into its advertising systems. Laurie Richardson, the company’s vice president of trust and safety, detailed in a blog post that the goal is to support the enforcement of advertising policies. C2PA support will start rolling out to Google’s ad and search systems in the coming months.
Further down the line, the company may also bring bring C2PA to YouTube. “We’re also exploring ways to relay C2PA information to viewers on YouTube when content is captured with a camera, and we’ll have more updates on that later in the year,” Richardson detailed.
The project is one of several that Google is working on in a bid to address the spread of AI-generated content. Last year, the company debuted SynthID, a tool that adds a watermark to images created with some of its AI models. The watermark is invisible and remains on the files to which it’s applied even if they’re modified or compressed.