OpenAI teams up with rival Google: ChatGPT fully integrates DeepMind's invisible watermarking, and launches an image authenticity verification tool.

AIMPACT News, May 20 (UTC+8), according to Dongcha Beating monitoring, OpenAI announced a partnership with its former rival, Google DeepMind. The company has fully integrated SynthID invisible watermarks into images generated by ChatGPT, Codex, and its API, and has officially become a compliant product with the C2PA image credential standard. It also launched a public-preview version of an image authenticity verification tool.

As an industry standard, C2PA metadata relies on cryptographic signatures to record provenance, but this metadata is easily lost when users download images, take screenshots, or convert image formats. To address this vulnerability, OpenAI introduced SynthID invisible watermark technology developed by Google DeepMind, directly embedding identifiers in image pixels that are invisible to the naked eye. Even if the image is subjected to secondary processing such as screenshots, SynthID can still persist. A two-layer defense mechanism combines the rich context of metadata with the durability of invisible watermarks.

Along with this update, OpenAI also launched an authenticity verification website (openai.com/research/verify), which is currently open to the public for preview. Users can upload images directly on the web page, and the tool will detect whether the images contain OpenAI’s Content Credentials credentials or SynthID invisible watermarks. However, OpenAI is very cautious in its detection conclusions. Because, under extreme circumstances, the watermark and metadata may still be forcibly removed, even if the tool finds no anti-forgery signals, it cannot decisively conclude that the image was not generated by AI.

(Source: BlockBeats)