OpenAI has announced new measures designed to make it easier to tell when an image has been created using its AI tools, as AI-generated visuals become harder to distinguish from real photos.
The company is adopting a global standard called C2PA, which embeds metadata inside image files to indicate whether they were generated by AI. This information acts like a digital label that helps verify an image’s origin.
In addition to metadata labeling, OpenAI is also working with Google to use a system called SynthID. This is an invisible watermark that is built into AI-generated images and is designed to remain detectable even if someone tries to edit or remove it.
Unlike basic metadata tags, SynthID is harder to erase because it stays embedded in the image itself, even after screenshots, cropping, or other changes. This makes it more reliable for tracking manipulated content.
OpenAI says these two methods are designed to work together. Metadata provides detailed information about how an image was created, while watermarks offer stronger protection against tampering.
The company also plans to release a public verification tool that allows users to upload an image and check whether it was generated by OpenAI’s systems. At launch, the tool will only work with OpenAI-created images, but the company hopes to expand it to other AI tools in the future.
The C2PA standard, developed by the Coalition for Content Provenance and Authenticity, is already used by several major tech companies, but adoption is still uneven across the industry. Because metadata can sometimes be altered, it works best when platforms agree to follow the same rules.
SynthID, developed by Google, is meant to address that weakness by staying intact even after image edits or transformations. Together, both systems aim to make AI image tracking more reliable and harder to bypass.
OpenAI says the combined approach is meant to strengthen transparency in AI-generated content and reduce confusion in online media as AI tools become more widely used.
