Skip to main content
Red Team Configurations allows users to create, edit, delete, and run preset configurations for AI Attack Simulation workflows. Configurations can include a target model, target system prompt, and severity mapping. Advanced options include applying a Runtime Project and execution strategy. Save reusable red team presets, so you can kick off evaluations without re-entering target models, system prompts, objectives, severity weighting, and execution limits each time. Start a run from a configuration to apply its defaults; per-run overrides always win.

​
Create Configuration

  1. In the Console, go to Attack Simulation > Configurations.
  2. Click Create Configuration. A configuration slide-out displays.
  3. Enter a name for the configuration.
  4. Optionally, enter a description.
  5. Select optional defaults:
    • Target Model: Select a model that is similar to what you have in your environment to simulate the attacks against your system prompt.
    • Target System Prompt: Enter the target system prompt.
  6. Optionally, you can override the Severity Mapping. You can change the default severity mapping to meet your organization’s requirements. Only changed severities are saved as overrides.
    • For the detection category you want to change the severity for, click the severity drop-down list, then select a different severity.
  7. Optionally, you can change the Advanced Settings to change the Runtime project settings and attacker technique parameters.
    • Runtime Project: Select a project to apply Runtime Policies to Interaction tagging. If no project is selected, the default project is used.
    • Execution Strategy: Select the number of additional random techniques. The minimum is one and the maximum is five.
    • Attacker Max Turns to Complete Objective: Set the maximum number of conversation turns allowed per technique when attempting to achieve an objective. The minimum is one and the maximum is five. The attack simulator will do multi-turn, trying to attack the target for N turns, and then go to the next session.
    • Sessions Per Technique: Set the number of independent sessions to run for each technique. The minimum is one and the maximum is five. This is the number of times you want to run the same technique or static prompt.
  8. Click Create Configuration.

​
Run Configuration

  1. In the Console, go to Attack Simulation > Configurations.
  2. For the configuration you want to run, click the Run button. The configuration slide-out displays.
    • If a target model and system prompt are not included in the configuration, you must add those before you can run the configuration.
  3. Click Start Evaluation.
  4. Go to Attack Simulation > Red Teaming to view the evaluation.