With AI image generators now widely accessible and more advanced than ever, distinguishing authentic images from fabricated ones has become increasingly difficult. However, on Tuesday, OpenAI introduced two new initiatives to help address this challenge. The company has adopted an open standard known as C2PA (Coalition for Content Provenance and Authenticity), which embeds a clear signal in metadata indicating that an image was created by AI. Additionally, OpenAI is collaborating with Google to integrate an invisible watermark called SynthID, which is harder to detect and more resistant to removal if malicious actors attempt to conceal their actions.

These new protections apply exclusively to images generated by OpenAI products. While they won't curb the flood of imagery from less reputable AI tools, they can help ensure that OpenAI does not contribute to the problem. OpenAI is also unveiling a public verification tool that will check for both signals, enabling users to easily determine whether an image was produced using AI. Initially, this tool will only work with images generated by OpenAI products, but the company hopes to expand its coverage to other tools over time.

Founded in 2021, the Coalition for Content Provenance and Authenticity (C2PA) is a nonprofit organization dedicated to mitigating the harmful effects of AI-generated imagery on public discourse. The C2PA standard has been adopted by several Google products, but industry-wide adoption remains inconsistent. Because the C2PA signal is clearly accessible in each file's metadata, it can be manipulated and is most effective among trusted users.

SynthID represents a newer effort designed to provide stronger resistance to tampering. Developed by Google, this watermark is engineered to persist even when bad actors attempt to remove it through screenshots, resizing, or digital manipulation. The two systems are intended to complement each other, with each addressing the other's weaknesses. "Watermarking can be more durable through transformations like screenshots, while metadata can provide more information than a watermark alone," OpenAI noted in its announcement. "Together, they make provenance more resilient than either layer would be on its own."