Content Moderation is an AI-powered service that uses cloud computing technology to identify and label inappropriate or non-compliant multimodal content. The service can detect risks in images, videos, text, and audio across various industries and business scenarios to improve moderation efficiency, enhance platform content quality, and optimize user experience.
Overview
Alibaba Cloud Content Moderation is an enterprise-grade intelligent content moderation service. It leverages Alibaba Group's extensive experience in content governance and advanced large model algorithms to automatically identify and accurately label inappropriate or non-compliant content across various media formats, including text, images, videos, and audio.
Core value
It helps you build an efficient content moderation system, significantly reducing manual review costs while increasing moderation speed and accuracy. This ensures your platform's content is compliant and safe, providing a better and more secure user experience.
Product positioning
Designed for content-driven businesses such as video websites, live streaming platforms, social media, e-commerce platforms, and online education, it provides a ready-to-use solution for identifying content risks, so you can focus on your core business development.
Mission and vision
The service aims to create a trustworthy digital content ecosystem by providing tools to ensure content safety and platform compliance, supporting the sustainable growth of the internet industry.
Core features
Alibaba Cloud Content Moderation provides three core functional modules that cover a complete range of use cases, including API integration, storage detection, and visual management. These modules cater to users with different technical skills and business needs.
Content Moderation API
Provides a standardized RESTful API that supports both real-time and asynchronous calls. Developers can quickly retrieve content detection results through simple HTTP requests and SDKs. The API also supports advanced features like batch processing and asynchronous callbacks, offering flexible integration with various business scenarios and technical architectures.
OSS compliance detection
Deeply integrated with Alibaba Cloud OSS, this feature allows you to automatically scan for and identify risks in content stored in a bucket. No coding is required—simply configure the settings on the console. It supports multiple modes, including incremental detection and scheduled detection, making it easier and more efficient to manage stored content.
Visual management console
Provides a powerful web-based management interface that supports data analytics, detection result queries, custom risk library management, and policy configuration. With intuitive charts and a streamlined workflow, the console makes content security management straightforward.
Version description and selection recommendations
The Content Moderation enhanced edition (Machine Moderation enhanced edition) is recommended over Machine Moderation V1.0. The enhanced edition provides more than 100 detection labels, compared with more than 40 in Machine Moderation V1.0, and can return multiple violation labels at the same time.
The enhanced edition includes preset business scenarios such as avatars and nicknames, and provides more flexible configuration.
For the same usage, the enhanced edition costs about 22% to 48% of Machine Moderation V1.0, and image moderation through OSS form upload costs even less.
Both editions are managed in the same Alibaba Cloud Management Console without separate sign-ins. To switch between editions, choose Machine Moderation enhanced edition or Machine Moderation V1.0 in the left-side navigation pane.
Workflow
Detection capabilities
The service supports content detection across multiple modalities and covers a wide range of common risk scenarios, providing comprehensive content security for your business.
Supported modalities
Identified risk categories
Pornography: Identifies pornographic, explicit, and sexually suggestive content to maintain a healthy platform environment.
Politically sensitive content: Detects politically sensitive topics and illegal or non-compliant speech to ensure content compliance and security.
Terrorism: Identifies content related to bloody violence, terrorism, and dangerous acts.
Disturbing or graphic content: Detects visually shocking, disgusting, or offensive content.
Ads and junk information: Identifies marketing promotions, fraudulent information, spam, and malicious traffic diversions.
Abuse and harassment: Detects personal attacks, discriminatory remarks, and malicious harassment.
Illegal and prohibited content: Identifies information related to gambling, drugs, controlled substances, and other illegal activities.
Personally Identifiable Information (PII) leakage: Detects sensitive personal data such as ID numbers, phone numbers, and bank card numbers.
AI-generated content (AIGC) detection: Determines whether text, images, audio, or video content is likely generated or synthesized by AI. This capability helps you manage AIGC on your platform.
Custom risk content: Content Moderation lets you create custom moderation agents that use the large model to detect unique risks.
Use cases
The service applies to various content-based platforms and business scenarios, providing tailored content security solutions for different industries.
Video websites
Performs pre-moderation and ongoing inspection of user-uploaded videos. Automatically identify non-compliant videos to reduce manual review workload and ensure platform compliance.
Live streaming platforms
Monitors live video streams and comments in real time. Quickly detect and handle policy violations to maintain a healthy streaming environment and improve user experience.
Social media platforms
Conducts comprehensive monitoring of user profiles, posts, private messages, and group chats to prevent the spread of harmful information and maintain a positive community atmosphere.
News and media
Reviews news articles and user comments to ensure information compliance, prevent the spread of inappropriate news, and enhance media credibility.
E-commerce platforms
Detects non-compliant information in product titles, descriptions, images, and reviews. Combat exaggerated advertising and the sale of prohibited items to regulate the marketplace.
Gaming and entertainment
Monitors in-game chats, nicknames, and guild information to create a healthy gaming environment and protect underage players.
Online education
Reviews course materials, homework submissions, and interactive discussions to ensure educational content is positive and meets regulatory requirements.
Logistics and transportation
Ensures compliance in order postings, user reviews, and private messages to secure business operations and optimize the platform environment.
FAQ
How can a Resource Access Management (RAM) user check whether the Alibaba Cloud account has activated the Content Moderation enhanced edition?
A RAM user cannot directly view the service status or activate the service. The RAM user must contact the owner of the Alibaba Cloud account. The account owner can log on to the Alibaba Cloud Management Console and visit the purchase or management page of the Content Moderation enhanced edition to check the status. If the service is not activated, the account owner must activate it.
Why does the console display Not Used after Content Moderation is activated? Is the service disabled if it is not configured or integrated?
A Content Moderation enhanced edition resource plan offsets charges for all enhanced services except human review, including OSS compliance detection and the Machine Moderation enhanced edition.
The service exists by default after activation, but its status changes from Not Used to Used only after a service is actually called for detection.
If no service is configured or called, detection does not run automatically and remains inactive. You can integrate only the services that you need, such as OSS compliance detection.
Does deploying Content Moderation in the Germany (Frankfurt) region involve cross-border data transfer?
For the small-model edition, data is processed locally in the Germany (Frankfurt) region, including computing, inference, and storage. No cross-border data transfer is involved.
For the large-model edition, data is sent to Singapore for centralized inference and storage, which involves cross-border data transfer.