Custom Agent Runtime Protection | Datadog

Custom Agent Runtime Protection

About This Program

Are you building AI agents in-house for customer-facing experiences, internal workflows, developer tooling, and more?

AI Guard helps secure these custom agents from runtime attacks, such as:

  • Prompt injection
  • Tool misuse
  • Data exfiltration
  • And many more OWASP Top 10 threats

AI Guard is built on Datadog’s deep observability foundation. It evaluates the full history of agent behavior, infrastructure signals, and data flows to discover unprotected agents, flag unsafe actions, and block attacks in real time. This gives you contextualized, high-fidelity protection that goes beyond AI model, cloud, and open-source guardrails.

Sign up for the Limited Availability paid program to integrate agentic runtime protection directly in your applications and secure your custom agents in real time, so you can ship AI fast without compromising security.

Related Resources

Interested in more of our latest features?

Help make the next releases of Datadog products our best yet.