AI Labeling Systems Face Make-or-Break Moment

We’re about to discover whether the tools designed to make deepfakes and AI-generated content easy to identify can actually deliver on their promise. SynthID and C2PA Content Credentials, two distinct technologies that invisibly tag image, video, and audio files with origin information, are undergoing their biggest expansion yet. This rollout offers a pivotal chance to push back against unlabeled AI fakery that’s currently deceiving people online.

At its I/O conference yesterday, Google announced that the ability to verify whether images carry SynthID markers , the invisible watermarking system applied to content from Google AI models , is now coming to Chrome and Search. That’s a major development because Chrome dominates the global browser and search market, putting AI verification tools directly in front of far more users. It also simplifies the checking process; previously, verifying an image for SynthID required uploading it to the Gemini app.

On top of that, Google’s verification interfaces will now also scan for C2PA information , provenance metadata embedded at the point of creation that reveals how a file was made, whether it was manipulated, and if AI tools were used. This C2PA adoption lets users check suspicious images from a single interface, rather than jumping between the Gemini app and dedicated C2PA verification portals, since files may carry only one type of label or none at all.

This kind of collaborative effort is exactly what we’ve been waiting for. Although both systems work differently, Google and the Content Authenticity Initiative (which promotes the C2PA standard) have made similar claims about what’s needed for success: universal adoption. That means more AI models must embed this data, and the online platforms where AI fakery is most frequently shared need to display that information clearly. For platforms that don’t check or present AI metadata, having verification tools built into the web browser could serve as a practical workaround.

OpenAI is also joining this expansion. Yesterday, it announced that it will now embed SynthID into images generated by ChatGPT, Codex, and the OpenAI API. The company already includes C2PA metadata in generated content, but I’ve observed that this data is often stripped out when posted to other platforms. OpenAI has also tried to temper expectations around C2PA, despite being a steering member of the standard and reaffirming its commitment. Here’s what OpenAI said on its C2PA help page before it was updated to include SynthID:

“Metadata like C2PA is not a silver bullet to address issues of provenance. It can easily be removed either accidentally or intentionally. For example, most social media platforms today remove metadata from uploaded images, and actions like taking a screenshot can also remove it. Therefore, an image lacking this metadata may or may not have been generated with ChatGPT or our API.”

For something considered the gold standard in content authenticity tech, that sounds remarkably fragile. Even Google calls C2PA the industry standard and pitches it to global governments as a solution for AI transparency and labeling requirements. But despite growing adoption by AI, hardware, and software providers, I rarely see it successfully used to verify AI fakery in the wild. SynthID seems more robust by comparison because it can’t be easily stripped out. For all its limited reach compared to C2PA, I can recall several instances where fact-checkers and media agencies have cited SynthID in debunking deepfakes online.

Both C2PA and SynthID can work together to cast a wider safety net. This isn’t an industry that would benefit from a verification standards war, but Google has a clear opportunity here to prove whether its system is more reliable and steal some of the spotlight that C2PA has claimed for itself. To prevent that, C2PA needs to demonstrate it can actually demystify where the content we see online originates.

That opportunity has already arrived. Google announced yesterday that Meta will start using C2PA metadata to tag images on Instagram that have been captured by a camera. Meta hasn’t responded to our questions about what this will look like or which cameras will be supported, though I suspect it will involve labels like “captured on Pixel 10,” similar to the “sent from my iPhone” notes in emails. This could help Instagram users distinguish “real” photos from convincing AI fakery, aligning with the future Instagram head Adam Mosseri has predicted: moving away “from assuming what we see is real by default.”

That is, if labeling actually works. Instagram already checks images for C2PA information, and its attempts to label AI-generated content have previously backfired after it applied AI labels to images that photographers insisted were their own work.

I wouldn’t rush to praise Google for this partnership either. The company preaches about the importance of AI transparency and fighting digital deepfakes, all while developing the very technology used to mislead people. It has positioned itself as both the supplier and the solution. I can forgive that if SynthID makes a noticeable difference in the fight against deepfakes, but I’m not holding my breath given the scale of the problem.

Robust or not, SynthID and C2PA can only detect watermarks if they’ve been added in the first place. I doubt many of the open-source models used to generate truly malicious deepfake content are lining up to adopt these systems. Provenance was never going to be a perfect solution, but now Google and C2PA have the chance to prove it’s not a complete waste of time.