Presenter: Arjoyita Roy, Tech Evangelist at A10 Networks 

Topic: The Missing Layer: Why AI needs Guardrails

Guardrails exist everywhere because speed without boundaries is dangerous. Yet as AI moves into production, one critical layer keeps getting skipped, which is security at the AI layer itself.

Most security strategies end at deployment. That's where the risk begins.

LLMs can't police themselves. A natural language instruction is not a security control. It works in testing, where inputs are predictable. Attackers don't send predictable inputs. And traditional web application firewalls, built for syntax, have no concept of intent or context. Interactions with AI systems like LLMs and AI agents are semantic and non-deterministic.

This session makes the case for a guardrail layer that understands language the way attackers use it.