We break it, we fix it, we write about it.

This session takes a defensive look at AI security, showing how teams can build an AI observability pipeline to trace prompts, responses, tool calls, and agent behaviours.

Gregory and Lancer will explore how this visibility can help detect prompt injection, suspicious tool use, memory poisoning, anomalies, and other emerging AI threats.

Attendees will also see how open-source technologies such as OpenTelemetry can be used to turn AI telemetry into actionable security alerts and strengthen applications over time.