With AI image generators becoming increasingly sophisticated, it's getting harder to distinguish between real and synthetic images. To tackle this issue, OpenAI has introduced two new measures: an open standard called C2PA and a SynthID watermark.

C2PA adds metadata to images indicating they were created by AI, while SynthID is designed to be harder to remove. Together, these should make it easier for users to verify if an image was generated using AI technology.

OpenAI has also developed a public verification tool that can check both signals, allowing anyone to test the authenticity of an image. While these measures only apply to images created by OpenAI products initially, the company hopes to expand their use in the future.

The Coalition for Content Provenance and Authenticity (C2PA) is a non-profit working to mitigate harmful effects from AI imagery on public discourse. Although C2PA has been adopted by Google products, consistent industry adoption remains a challenge.

Both systems are complementary, with metadata providing more information than a watermark alone, while watermarks can be more durable through transformations like screenshots. Together, they aim to make provenance more resilient against manipulations.