OpenAI says it’s getting serious about AI detection and labeling

OpenAI is enhancing its AI content identification efforts by combining C2PA metadata with Google’s SynthID watermarking technology. This multi-layered approach is designed to increase the resilience of provenance signals, as metadata is often stripped during file transfers or screenshots, whereas watermarks provide a more durable verification layer. Initially applied to images created by ChatGPT, Codex, and the OpenAI API, the company is also launching a public portal for users to verify these signals. While OpenAI acknowledges that no detection method is foolproof, this strategy aims to close existing gaps in identifying AI-generated content and deepfakes online.

(Source：The Verge)