Sweet AI Security Platform (AISP) provides security coverage across the AI lifecycle, from model layer to agent execution in production environments. The platform addresses security challenges across foundation models, orchestration frameworks, GPUs, embeddings, and autonomous AI agents. The platform includes AI-BOM (AI Bill of Materials) capabilities that track models, dependencies, versions, and risk across the AI ecosystem. AI-SPM (AI Security Posture Management) continuously monitors AI components for misconfigurations, exposed endpoints, vulnerabilities, and policy violations. For runtime protection, AISP discovers and monitors AI agents, including shadow or unmanaged instances. It traces agent actions across timelines and workflows, assesses architectural risk, and manages permissions with minimal privilege enforcement. The platform calculates blast radius for potential attacks on agents and enables policy enforcement. AISP includes red teaming capabilities to test agent behavior under adversarial conditions. The AIDR (AI Detection and Response) component detects and blocks attacks such as prompt injections and hallucinations by routing AI agent traffic through an AI Gateway for prompt analysis and malicious operation blocking. The platform establishes behavioral baselines for AI agents to detect deviations and unexpected workflows. The platform supports compliance frameworks including OWASP, SOC 2, ISO, and NIST standards.