Securing the AI Workload: Defending Enterprise LLMs and Agentic Systems

Abstract

As enterprises move Generative AI and LLMs into production, security teams face a paradigm shift. Securing AI is no longer just about defending against "AI-powered attackers" (AI as a Weapon); it is fundamentally about safeguarding the AI infrastructure itself (AI as a Workload).

This session explores the unique security challenges of integrating LLMs into corporate environments. We will focus on critical threat vectors like prompt injection, sensitive data leakage via RAG pipelines, and the governance risks of autonomous agentic systems.

As an F5 Solutions Architect, Sven Müller will deliver a practical corporate defense blueprint using the F5 AI Security & Delivery Platform (ADSP). Attendees will learn how to:

Enforce real-time In-Line AI Guardrails to sanitize prompts and prevent outbound data leakage.

Leverage Continuous Agentic Red Teaming for proactive model benchmarking and auto-remediation.

Secure LLM APIs and manage token consumption at the network ingestion layer.

Gain the strategic framework and technical insights needed to confidently secure your enterprise AI adoption.