Boosting AI Content Transparency with C2PA

As we continue to integrate AI into more products and services to boost creativity and productivity, we're committed to helping users understand how content is created and modified over time. We believe it's essential for people to have access to this information, which is why we're investing in tools and innovative solutions like SynthID.

We also recognize the importance of collaborating with others in the industry to enhance overall transparency as content moves across different platforms. That's why we joined the Coalition for Content Provenance and Authenticity (C2PA) as a steering committee member earlier this year.

Today, we're excited to share updates on our efforts to develop the latest C2PA provenance technology and integrate it into our products.

Advancing existing technology to create more secure credentials

Provenance technology helps clarify whether a photo was taken with a camera, edited by software, or generated by AI. This information empowers our users to make more informed decisions about the content they engage with—be it photos, videos, or audio—and fosters media literacy and trust.

As a steering committee member of the C2PA, we've collaborated with other members to enhance the technology that attaches provenance information to content. Throughout the first half of this year, Google contributed to the development of version 2.1 of the Content Credentials technical standard. This new version offers stronger protection against a broader range of tampering attacks, thanks to stricter technical requirements for validating the content's provenance history. By bolstering these protections, we ensure that the attached data remains accurate and untampered.

Incorporating the C2PA’s standard into our products

In the coming months, we'll integrate this latest version of Content Credentials into several of our key products:

• Search: If an image includes C2PA metadata, users can use our "About this image" feature to check if it was created or edited with AI tools. "About this image" provides context about online images and is available in Google Images, Lens, and Circle to Search.
• Ads: We're beginning to integrate C2PA metadata into our ad systems. Our aim is to expand this integration over time and use C2PA signals to help enforce key policies.

We're also exploring ways to display C2PA information to viewers on YouTube when content is captured with a camera, and we'll share more updates on this later in the year.

We'll ensure that our implementations validate content against the forthcoming C2PA Trust list, which helps platforms confirm the content's origin. For instance, if data indicates an image was taken by a specific camera model, the trust list helps verify the accuracy of this information.

These are just a few examples of how we're implementing content provenance technology today, and we'll continue to expand its use across more products over time.

Continuing to partner with others in the industry

Establishing and signaling content provenance is a complex challenge, with various considerations depending on the product or service. While there's no one-size-fits-all solution for all online content, collaborating with others in the industry is crucial for creating sustainable and interoperable solutions. That's why we're encouraging more services and hardware providers to adopt the C2PA’s Content Credentials.

Our work with the C2PA aligns with our broader approach to transparency and the responsible development of AI. For example, we're expanding the use of SynthID—Google DeepMind's embedded watermarking—to additional generative AI tools for content creation and various media types, including text, audio, visual, and video. We've also joined several other coalitions and groups focused on AI safety and research, introduced a Secure AI Framework (SAIF) and coalition, and continue to make progress on the voluntary commitments we made at the White House last year.