AI Agents Pose New Cybersecurity Risks Radware Report Warns

A new report from Radware highlights the growing risks posed by AI agents in the cybersecurity landscape. These agents, powered by large language models (LLMs), can act autonomously, access tools and resources, and interoperate with other agents, creating a new class of threats.

Key Findings:

• Autonomous AI Agents: Organizations are deploying LLM-powered agents capable of reasoning, invoking tools, and communicating with other agents through emerging protocols like Model Context Protocol (MCP) and Agent-to-Agent (A2A). These systems create transitive chains of access to enterprise resources, which are difficult to monitor and secure with traditional methods.

• New Attack Surfaces: The capabilities of AI agents introduce risks such as prompt injection, tool poisoning, and lateral compromise. Malicious AI platforms, like subscription-based XanthoroxAI, are lowering the barrier for cybercrime by offering automated attack toolkits.

• Faster Exploit Development: LLMs like GPT-4 can generate working exploits from vulnerability descriptions faster than human researchers, shortening the window for defenders to respond to newly disclosed vulnerabilities.

Pascal Geenens, Director of Cyber Threat Intelligence at Radware, emphasized the urgency of addressing these risks: "We are not entering an AI future; we are already living in it. AI is no longer just a tool; it is a participant in systems, a co-author of code, a decision-maker, and increasingly, an adversary."

Call to Action:

Radware will host a webinar on September 25th at 11am ET to discuss the findings and strategies for securing AI-driven infrastructures.

Image credit: Napong Rattanaraktiya/Dreamstime.com