Radware has launched its Agentic AI Protection Solution, expanding the Radware Platform to address the rapidly emerging risks associated with autonomous AI agents.
As enterprises increasingly deploy agentic AI to automate decisions and improve productivity, traditional AI security guardrails—designed primarily for governance—are proving insufficient. Radware’s new solution introduces agentic security posture management, using patent-pending, automated behavioral analysis to detect and mitigate malicious activity in real time.
The solution is designed to protect against a wide range of agent-specific threats, including prompt injection attacks, tool misuse, unauthorized data access, and human–agent trust exploitation. It provides real-time discovery of AI agents, intent-based runtime security, deep integration with leading agent platforms such as Microsoft Copilot and AWS Bedrock, and continuous risk assessment through a dynamic AI security posture map.
Radware’s launch follows the company’s recent identification of ZombieAgent, a zero-click indirect prompt injection vulnerability that enables persistent, covert data exfiltration by compromising an AI agent’s memory—highlighting a growing blind spot in agentic AI environments.
Aligned with the OWASP Top 10 for Agentic AI and leveraging the AI Vulnerability Scoring System (AIVSS), Radware Agentic AI Protection aims to help enterprises adopt autonomous AI securely while maintaining control over sensitive data and critical systems.