OpenAI Teams Up with Google: ChatGPT Fully Integrates DeepMind's Steganography Watermark and Introduces Image Authenticity Verification Tool

According to BotBot Beating monitoring, OpenAI has announced a collaboration with its former rival Google DeepMind to integrate SynthID invisible watermarks into images generated by ChatGPT, Codex, and the API. This integration makes OpenAI a fully compliant C2PA Image Verifiable Claims standard product and has launched a public preview of an image authenticity verification tool.

As the industry-standard C2PA metadata relies on cryptographic signatures to track image sources, this metadata is highly vulnerable to being lost when users download, screenshot, or convert the image format.

To address this vulnerability, OpenAI has introduced SynthID invisible watermarking technology developed by Google DeepMind, which embeds imperceptible identifiers directly into the image pixels.

Even after secondary processes like screenshots, the SynthID watermark remains intact. This dual-layer defense mechanism combines rich metadata context with the persistence of invisible watermarks.

In conjunction with this update, OpenAI has also launched an official image verification website (openai.com/research/verify), which is currently available for public preview.

Users can directly upload images on the website to check for OpenAI's Content Credentials or SynthID invisible watermarks.

However, OpenAI is very cautious in its verification conclusions. Due to the possibility of forcefully stripping off watermarks and metadata in extreme cases, even if the tool does not detect any anti-counterfeiting signals, it cannot definitively conclude that the image was not generated by AI.