AI technology is advancing rapidly, creating a surge of interest in both social and commercial sectors. While this generates a vast amount of engaging content, it also makes distinguishing authentic content from AIGC increasingly difficult. According to the Administrative Measures for Labeling AI-Generated Synthetic Content, providers of online information services must detect and label AIGC to inform the public of its synthetic origin. This document describes the AIGC detection and labeling services and best practices provided by Alibaba Cloud Content Moderation. This approach is based on our solution, which was selected for the National Cybersecurity Standard Application Practice Case Library.
Solution overview
At a sub-forum of the TC260 Security Standards Week, TC260 recognized Alibaba Cloud's solution as an AI Security Standard Application Practice Case. The solution demonstrates the application of the GB 45438-2025 standard, "Cybersecurity Technology: Methods for Labeling AI-Generated Synthetic Content," within Alibaba Cloud's end-to-end technical framework for detection, digital watermarking, and verification.
In this practice case, Alibaba Cloud built a three-step compliance framework: AIGC detection, digital watermarking, and compliance verification. Refined through its application in Qwen and other Alibaba products, this framework provides a proven model for other platforms.
Prerequisites
Activate Content Moderation Enhanced Edition.
Integrate the image moderation, text moderation, voice moderation, and video moderation services of Content Moderation Enhanced Edition.
AIGC detection services
Alibaba Cloud Content Moderation provides capabilities to detect AIGC across images, text, audio, and video. The following table lists the available services.
Modality | Service | Detection scope | Use cases |
Image moderation | Service: AI-Generated Image Detection Service: aigcDetector | Object: Image. Scope: Detects whether an image is likely generated by AIGC. | Use this service to determine if an image is likely generated by AIGC. For more information, see AIGC detection for image moderation. |
Service: AI-Generated Image Detection with Implicit Identifier Service: aigcDetectorFull | Object: Image. Scope: Detects whether an image is likely generated by AIGC and whether it contains an AI-generated implicit identifier. If an implicit identifier is found, the service returns the AIGC metadata of the image. | Use this service to determine if an image is generated by AIGC and to detect its implicit identifier and AI generation traces. For more information, see AIGC detection for image moderation. | |
Service: AI-Generated Image Detection for Video Frames Service: aigcDetectorForFrame | Object: Video frame. Scope: Detects whether a video frame is likely generated by AIGC. | Use this service to determine if a video frame is generated by AIGC. For more information, see AIGC detection for image moderation. | |
Service: AI-Generated Image Detection - Pro Service: aigcDetector_pro | Object: Image. Scope: Detects whether an image is likely generated by AIGC or is a suspected synthesis (including face synthesis). | Use this service to detect forged images. For more information, see AIGC detection for image moderation. | |
Service: AI-Generated Image Detection - Ultra Service: aigcDetector_ultra | Object: Image. Scope: Detects whether an image is likely generated by AIGC, is a suspected synthesis (including face synthesis), or has undergone suspected AI-based partial editing. | Use this service to detect forged images, especially those that have been partially edited by AI. For more information, see AIGC detection for image moderation. | |
Text moderation | Service: AI-Generated Text Detection Service: text_aigc_detector | Object: Text. Scope: Detects whether text is likely generated by AIGC. | Use this service to detect and label AI-generated text. For more information, see AIGC detection for text moderation. |
Voice moderation | Service: AI-Generated Voice Detection Service: voice_aigc_detector | Object: Audio. Scope: Detects whether audio is likely generated by AIGC. | Use this service to detect and label AI-generated audio. For more information, see AIGC detection for voice moderation. |
Video moderation | Service: AI-Generated Video Detection Service: videoAigcDetector | Object: Video. Scope: Detects whether a video is likely generated by AIGC and whether it contains an AI-generated implicit identifier. If an implicit identifier is found, the service returns the AIGC metadata of the video. | Use this service to detect and label AI-generated video. For more information, see AIGC detection for video moderation. |
Required APIs for integration and configuration:
Modality | API | Service configuration |
AIGC detection for images | aigcDetector aigcDetectorFull aigcDetectorForFrame aigcDetector_pro aigcDetector_ultra | |
AIGC detection for text | text_aigc_detector | |
AIGC detection for voice | voice_aigc_detector | |
AIGC detection for video | videoAigcDetector |
Workflows
Image publishing
When publishing images, sharing assets, or processing image-based insurance claims on a platform, use the AIGC image detection service. This service checks if an image is likely generated by AI, contains an implicit identifier, or has undergone AI-based partial editing. You can then label or flag the AI-generated images accordingly.
Select an AIGC image detection service:
Content Moderation workflow overview:
Required API for integration and configuration:
Step | API | Service configuration |
Step 2 | aigcDetectorFull |
Video publishing
When uploading or publishing videos on your platform, use the AIGC video detection service. This service checks if a video is likely generated by AI. You can then label or flag the AI-generated video accordingly.
Content Moderation workflow overview:
Required APIs for integration and configuration:
Step | API | Service configuration |
Step 2 | videoAigcDetector | |
Step 5 | Service configuration in the console |
Voice publishing
When uploading or publishing voice content on your platform, use the AIGC voice detection service. This service checks if a voice clip is likely generated by AI. You can then label or flag the AI-generated voice content accordingly.
Content Moderation workflow overview:
Required API for integration and configuration:
Step | API | Service configuration |
Step 2 | voice_aigc_detector |
Text publishing
When publishing text on your platform, use the AIGC text detection service. This service checks if the text is likely generated by AI. You can then label or flag the AI-generated text accordingly.
Content Moderation workflow overview:
Required API for integration and configuration:
Step | API | Service configuration |
Step 2 | text_aigc_detector |