Zscaler AI Guard is a runtime security service for AI traffic, built on Zscaler's inline inspection platform. It enforces policy on prompts and responses exchanged between users or applications and large language models (LLMs). More than eighteen specialised detectors with multilingual support cover, among other things, prompt injection, jailbreak attempts, malicious URLs, exposed secrets, personal data, toxicity and mentions of competitors. Applications call the service through an API; integrations exist for coding assistants, CI/CD pipelines and gateways. Each scan returns allow, detect (log only) or block.
AI Guard is a commercial, closed-source cloud service from Zscaler and requires a subscription with API access. The public sources consulted do not state in which regions prompts and responses are processed or retained. Because this content is often business-sensitive, a European organisation should obtain written confirmation before adoption of where inspection takes place, whether and for how long content is stored, and under which jurisdiction the vendor operates.