Daxa AI Factory Security Platform is a security platform designed to protect AI factories, agentic workflows, and enterprise AI deployments. It operates on a "shift-left" model, applying data access controls and compliance checks before data reaches large language models (LLMs), rather than filtering outputs after the fact. The platform uses a TwinGuard architecture consisting of two core components: - SafeConnectors: Pull fine-grained permissions from enterprise systems - SafeRetriever: Applies those permissions before any data reaches the LLM, ensuring only authorized and compliant context is forwarded It includes Pebblo MCP, an MCP-native security layer that governs data and tool access at the protocol level by validating permissions and sanitizing payloads before they reach AI assistants. Core threat vectors addressed include: - Secrets and IP leaks to external models - Data overreach beyond role or project context - Autonomous agents making unsafe or unsupervised changes - Prompt injection attacks - Agent compromise via over-permissioned autonomous agents - Data and model poisoning through manipulated ingestion or training pipelines - Inference and extraction attacks at the model serving layer - Supply chain risk from third-party MCP servers The platform targets enterprises running AI coding assistants and agentic workflows, providing real-time data loss prevention, RBAC-aligned access controls, agent behavior guardrails, and injection/supply-chain defenses without requiring developer workflow changes. It is recognized in the 2025 Gartner Market Guide for AI Trust, Risk and Security Management (AI TRiSM).