Alibaba Cloud AI Guardrails

Alibaba Cloud AI Guardrails is a cloud-based AI security service that monitors and moderates inputs and AI-generated content (AIGC) in real time. It is designed to protect AI applications from misuse, data leakage, and adversarial attacks by applying a three-layer protection architecture: Content Moderation, Dynamic Defense, and Model Health.

The service is powered by Alibaba Cloud Tongyi Qianwen and uses multimodal detection, threat intelligence analysis, and an intelligent policy engine.

Key capabilities include detecting and blocking sensitive or privacy-related content (e.g., personal identity data, corporate trade secrets) to prevent data leakage in compliance with regulations such as GDPR and PIPL. It performs multi-dimensional compliance checks on generative AI inputs and outputs, covering politically sensitive content, pornography, bias, discrimination, and harmful values.

It also detects and defends against prompt injection attacks, including jailbreak prompts, role-play inducements, and system command tampering.

The platform offers a configurable console where users can enable or disable specific risk checks and create custom detection templates. It supports online testing of moderation policies, result querying, and risk reporting. Integration is available via API and one-click activation through Alibaba Cloud Model Studio.

Billing is offered on a pay-per-call or token-based model, with Text Moderation and Sensitive Content Moderation priced at USD 0.0006 per use. Prompt injection attack moderation is currently in public preview with a free trial available.