Zscaler AI-SPM (AI Security Posture Management) provides visibility and security controls for AI models, agents, and services deployed in cloud environments. The product discovers and inventories AI-related services, datasets, models, and vectors across managed platforms like Amazon Bedrock, Microsoft Azure Foundry AI, and Google Vertex AI, as well as unmanaged services like Hugging Face and Ollama. The platform uses LLM classification to automatically discover, classify, and assess risks of sensitive data connected to AI services. It identifies misconfigurations, excessive permissions, vulnerabilities, and data exposure risks across AI deployments and retrieval-augmented generation (RAG) frameworks. The system correlates threats to determine hidden attack paths and provides risk prioritization based on likelihood and impact. AI-SPM monitors data flows and access to sensitive data used in AI training datasets, alerting on critical and regulated data usage. It analyzes prompt and output logs to detect model misuse and potential data exposure. The product provides guided remediation for security issues with step-by-step instructions and complete context. The platform includes compliance monitoring for standards such as NIST AI RMF 600-1, EU AI Act, HIPAA, and GDPR through continuous monitoring and reporting. It integrates natively with the Zscaler Data Security platform and can connect with DSPM/DLP solutions and ITSM tools. The system provides model inventory with information on publisher, country of origin, licensing terms, and risk factors.