China: TC260 adopted Guidelines on Identification Methods for AI-Generated Synthetic Image Content

On 28 August 2025, the National Technical Committee on Cybersecurity Standardisation (TC260) adopted the Cybersecurity Standard Practice Guide – Artificial Intelligence Generated Synthetic Content Identification Method: File Metadata Implicit Identification for Image Files, which was formulated to implement the requirements of the mandatory national standard GB 45438-2025 Cybersecurity Technology — Identification Methods for AI-Generated Synthetic Content. The Guide specifies implicit metadata labelling methods for AI-generated and synthesised image files across formats including HEIF/HEIC, JPEG/JPG, PNG, TIFF, WebP, and GIF, establishing a unified technical basis for content traceability. It introduces a general scheme based on the Extensible Metadata Platform (XMP) standard using the TC260 namespace (http://www.tc260.org.cn/ns/AIGC/1.0/) and prescribes standardised fields such as Label, ContentProducer, ProduceID, ReservedCode1, ContentPropagator, PropagateID, and ReservedCode2. The Guide further provides detailed instructions for embedding metadata into each format’s structure, such as APP1/XMP fields for JPEG, iTXt/XMP fields for PNG, IFD0 tags for TIFF, Application Extension XMP packets for GIF, and equivalent fields for WebP and HEIF/HEIC in reference to ISO/IEC 23008-12. In addition, the annexes present implementation examples, including a Python-based code sample for metadata insertion, and alternative schemes based on EXIF UserComment fields or PNG tEXt chunks, thereby supporting AI content generation and network information service providers in fulfilling their obligations for implicit labelling of synthetic image content.