OpenAI adds C2PA and SynthID to verify AI-generated images

With AI image generators widely available online, distinguishing authentic images from AI-generated ones has become more difficult. On Tuesday, OpenAI announced two new measures to help address this problem. The company is implementing metadata from the Coalition for Content Provenance and Authenticity (C2PA) alongside SynthID, a Google-developed watermarking technology. However, these new protections only apply to images generated by OpenAI products. Because these protections are limited to OpenAI’s own models, they will not affect imagery coming from other AI tools. Instead, they are intended to help ensure that OpenAI is not contributing to the problem of unverified AI-generated content.

To help users check for these signals, OpenAI is previewing a public verification tool that will allow users to easily test whether an image was generated using AI. Initially, this tool will only check for C2PA and SynthID signals on images generated by OpenAI’s own models, though the company hopes to expand its capabilities to cover other tools over time.

The two systems are designed to work together to address the limitations of each individual method. The Coalition for Content Provenance and Authenticity (C2PA) is a non-profit founded in 2021. While C2PA provides an open standard that embeds clear signals directly into an image’s metadata, its adoption remains inconsistent across the industry, even though it has been adopted by a range of Google products. Because the C2PA signal is clearly accessible within the metadata of each file, it can be manipulated, making the standard most useful among trusted users.

To counter this vulnerability, OpenAI is partnering with Google to implement SynthID. This watermarking technology is designed to be more robust against digital manipulation. The watermark remains detectable even when bad actors attempt to remove it through screenshots, resizing, or other alterations. By combining both approaches, the company aims to build a more secure verification system. As OpenAI noted in its announcement, “Watermarking can be more durable through transformations like screenshots, while metadata can provide more information than a watermark alone, Together, they make provenance more resilient than either layer would be on its own.”

Why it matters

This move addresses the growing difficulty of distinguishing authentic images from AI-generated ones by combining metadata and watermarking to create a more resilient provenance standard.