AIShield Guardian is a post-deployment security platform for AI/ML models and Large Language Model (LLM) applications. It operates as both an ML firewall and a GenAI guardrails system, providing real-time threat detection and mitigation for deployed AI systems. For ML applications, Guardian detects and mitigates adversarial attacks, data poisoning, and model extraction attempts. It provides real-time intrusion detection and telemetry across cloud and edge environments. For Generative AI and LLM applications, Guardian enforces over 75 configurable guardrails covering prompt injection prevention, jailbreak detection, invisible text detection, PII anonymization, bias detection, and dynamic redaction of sensitive data. Guardian supports two operational modes: Audit Mode (Out-of-Band), which monitors traffic without interfering with the target LLM, and Block Mode (In-Sync), which actively filters and orchestrates traffic in real time. Additional capabilities include: - Advanced content analysis and filtering to prevent code injection and sensitive data exposure - Data integrity validation for input/output pipelines - Multi-modal support for text and image-based inputs/outputs - Compliance reporting with pre-configured templates aligned to OWASP Top 10 for ML/LLM, NIST standards, and MITRE ATLAS frameworks - Protection for agentic AI workflows via integration with AI orchestration frameworks - Deployment flexibility across SaaS, cloud, hybrid, and on-premises environments (Docker, Kubernetes) - Automatic updates for SaaS and monthly updates for self-hosted deployments with zero-downtime upgrades