Agent Frameworks

Agent frameworks are software toolkits, such as LangChain, AutoGen, and CrewAI, that let developers build AI systems which take actions on their own. These systems can call tools, read and write files, query databases, and chain several steps together with limited human review. For compliance and risk teams, this matters because an agent built on a framework can reach internal systems and act on company data, so its permissions, logging, and oversight need to be governed as carefully as any employee's access.

1 item

ResearchGlobal2026-10-27

Four Agent Framework Flaws Let Attackers Cross Session Boundaries and Bypass Authorization

Security researcher Corgea has published advisories for four widely used AI agent and infrastructure frameworks: Strands Agents SDK, LangGraph, FastMCP, and MLflow. The flaws allow attackers to access other users' session data and navigate outside permitted file paths. They can also bypass access controls after data is processed and make unauthorized outbound network requests from scoring components. Organizations running these frameworks in production should treat the advisories as requiring immediate dependency review.

agentic AI agent frameworks authorization bypass dependency security session isolation LangGraph MLflow FastMCP