Trust but verify: A security playbook for agentic AI in production
You’ll learn:
- Why AI agents create a new attack surface that traditional security controls weren’t built to see
- How leading teams are governing agents in production — mapping the threat model and building in guardrails before agents scale
- The role of full-context visibility in detecting prompt injection, jailbreaks, and data exfiltration
Watch Now
As organizations adopt AI agents, they introduce new attack vectors beyond infrastructure, including prompts, model context, and sensitive data flowing through telemetry between developer environments and cloud services. Security teams must balance innovation with control, as agents without proper safeguards can operate outside expected boundaries and introduce risk.
In this session, we'll discuss how leading security teams are building a repeatable playbook for governing agents in production. As organizations race to deploy, the teams staying ahead are the ones treating agents as a first-class attack surface.
Event Speakers
Emmanuelle Lejeail is an Engineering Manager at Datadog, where she leads the AI Guard team. She has worked in security for four years, originally in application security and now in AI security, building tools that detect and block attacks against LLM applications in real time.
Alexa Levine is a Security Product Marketing Specialist at Datadog, where she focuses on AI Guard, Datadog’s protection layer for securing AI-powered applications and agents. She works at the intersection of security and go-to-market, helping teams understand how to safeguard their AI systems against emerging threats.
Mike Shema hosts the Application Security Weekly podcast.