# Red Team

**Red Team** is the scope-builder for a Wallarm professional-services red-team engagement against your AI estate. It is not a self-serve testing tool. The view sits in the AI Hypervisor UI so the engagement can be scoped against the same asset inventory the platform has already discovered ([Registry](https://docs.wallarm.com/ai-hypervisor/registry.md), [Findings](https://docs.wallarm.com/ai-hypervisor/findings.md)). Use **Red Team** when you need: * **Pre-deployment validation** of a new AI feature against jailbreak and prompt-injection patterns specific to your business logic. * **Pre-audit assurance:** an independent adversarial assessment to attach to an EU AI Act or SOC 2 evidence pack. * **Coverage-gap testing:** confirm that the detections and PII rules you have actually fire under realistic attacker pressure. * **Tool or MCP misuse scenarios:** assess whether tools exposed to agents can be coerced into out-of-scope actions, with the platform-recorded traces available as evidence afterwards. For day-to-day signal on agents misbehaving in production, work from [User Tracks](https://docs.wallarm.com/ai-hypervisor/user-tracks.md) and [Findings](https://docs.wallarm.com/ai-hypervisor/findings.md). **Red Team** is for scheduling a paid engagement, not for real-time control.