FrontRow

AI Is Now Part of Your Attack Surface. Are You Testing It?

Secarma explains how chatbots, copilots and AI agents get security tested.

Recorded · added 3 October 2026
Online
RecordingArtificial Intelligence & CopilotCybersecurityBusiness Apps & AutomationFree

About this session

Once a chatbot, copilot or agent is connected to internal knowledge, customer data, APIs or the ability to take actions, it becomes part of your attack surface. This recorded session from Secarma explains what AI security testing actually involves, and why testing the model on its own isn't enough.

It covers how prompt injection and manipulated inputs can affect connected systems, where sensitive information can leak through AI-powered applications, how testers check what a chatbot or agent can access, retrieve and reveal, how permissions and human approval controls are tested, and how AI testing fits alongside application, API and infrastructure penetration testing.

It finishes with what to consider testing before launching or expanding an AI feature. I've attended plenty of Secarma's sessions, and they consistently explain things properly rather than selling at you.

What you’ll learn

What a proper AI security test covers, from prompt injection to what your chatbot or agent can access

Who it’s for

Organisations launching or expanding chatbots, copilots or AI agents that connect to internal data, business applications or APIs, and the security and development teams responsible for them.

Keep going

Something wrong with this listing? Tell us and we will put it right.