OpenAI adds C2PA metadata and SynthID watermarks to AI images

OpenAI has announced two new events designed to help the public determine whether an image was generated by its artificial intelligence models. The company is officially joining the Coalition for Content Provenance and Authenticity (C2PA), simultaneously partnering with Google to place an invisible SynthID watermark on OpenAI’s image outputs.

The moves are a meaningful step towards transparency AI generated imagesalthough their scope is limited to content produced by OpenAI’s own tools.

Two systems, one goal

Established in 2021 by Adobe, Arm, BBC, Intel, Microsoft and Truepic, the C2PA standard adds metadata to a file that records its origin and any edits made along the way. It has since been ratified as an ISO standard and adopted by a number of Google products, although adoption remains patchy in the wider industry. Because the C2PA signal resides in the file’s metadata, it is publicly accessible, meaning it can be removed or manipulated. The standard is most reliable among trusted users and platforms that choose to protect it.

SynthID, developed by Google DeepMind, takes a different approach. Instead of adding readable metadata, it places an invisible watermark directly on the image. This watermark is designed to persist even through screenshots, resizing, compression, and other forms of digital manipulation, making it difficult for bad actors to root out.

How OpenAI Origin Works, source: OpenAI

The two systems are designed to complement each other. As OpenAI explains, watermarking offers persistence through changes such as screenshots, while metadata provides richer contextual information than a watermark alone. Together, the company claims, they create a more robust provenance system than either layer would have independently.

Public inspection tool with warnings

Along with the announcement, OpenAI is looking at a public verification tool that verifies both C2PA credentials and the SynthID watermark. The tool will allow anyone to upload an image and determine if it was generated by one of the OpenAI models.

For now, the tool only covers images produced by OpenAI products, though the company said it hopes to expand its scope over time. This is a significant limitation. The flood AI generated images online circulation stems from an ecosystem of multiple tools, many of which have little incentive to adopt provenance standards. OpenAI’s new measures may help ensure the company isn’t contributing to the problem, but they won’t do anything to address images from less careful sources.

Part of a wider push

The announcement comes amid growing concern from governments and civil society about the role of AI-generated content. disinformation and public debate. C2PA attracted more than 6,000 members and affiliates by early 2026, and its specification reached version 2.1 last year. OpenAI has now joined the coalition’s steering committee, placing it alongside Adobe, Microsoft and other founding members in shaping the future direction of the standard.

Google, in turn, is expanding the use of SynthID in its products. The partnership with OpenAI marks the first time the technology will be incorporated into the results of a major competitor, a notable example of cross-industry collaboration on artificial intelligence. security and transparency.

Again, the practical impact of these measures depends on how widely they are applied beyond the companies already at the table. AI-generated content detection remains a cat-and-mouse problem, and origin signals are only as useful as the platforms willing to verify them. OpenAI’s two-tiered approach has a reasonable technical basis, but what’s harder for the rest of the industry to follow is one that no one company can solve alone.