DeepKeep Maps the AI Agent Attack Surface

DeepKeep has launched an AI agent attack surface scanning and discovery solution designed to help enterprises understand and manage the risks introduced by agentic AI. The new AI Agent Scanner maps what each AI agent can access, which tools and data sources it connects to, and where potential vulnerabilities may exist across enterprise workflows.

As organizations move from basic chatbots to LLM-based agents that can interact with business applications, cloud services, and other systems, their attack surface expands. These agents can access finance platforms, collaboration tools, operational systems, and external data sources. That connectivity creates new exposure points that traditional security tools were not built to address.

The scanner performs attack surface mapping across agentic workflows, identifying connected tools, data sources, and possible weaknesses. It generates a visual risk map aligned with the OWASP Top 10 for Agentic Applications, helping security teams see how components may be exploited and what controls are needed. The goal is to provide visibility during both development and production.

DeepKeep also includes runtime protection for select agentic frameworks. The platform identifies where AI firewalls and guardrails should be placed based on observed behavior and access patterns. The solution supports major agentic frameworks, including Microsoft-based frameworks, Agentforce, OpenAI Agents, CrewAI, Amazon Bedrock AgentCore, n8n, and Make. The company plans to expand its agentic AI security capabilities further in 2026.