Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges

TL;DR

This survey explores the unique security threats posed by agentic AI systems with LLMs, reviews evaluation methods, and discusses defense strategies to promote secure development.

Contribution

It provides a comprehensive taxonomy of agentic AI security threats, reviews recent benchmarks, and discusses technical and governance defense strategies.

Findings

Identifies specific security risks of agentic AI systems.

Reviews current benchmarks and evaluation methodologies.

Highlights open challenges in securing agentic AI.

Abstract

Agentic AI systems powered by large language models (LLMs) and endowed with planning, tool use, memory, and autonomy, are emerging as powerful, flexible platforms for automation. Their ability to autonomously execute tasks across web, software, and physical environments creates new and amplified security risks, distinct from both traditional AI safety and conventional software security. This survey outlines a taxonomy of threats specific to agentic AI, reviews recent benchmarks and evaluation methodologies, and discusses defense strategies from both technical and governance perspectives. We synthesize current research and highlight open challenges, aiming to support the development of secure-by-design agent systems.