What are Offensive AI Ranges?
Offensive AI Ranges let your offensive and red teams practice using AI and AI agents in their engagements. Instead of learning about these tools in theory, your team works in a realistic range with AI built in, so they experience how the tools behave during a live engagement.
You learn how different AI models and harnesses behave, how to use them effectively in offensive work, where they go wrong, how model guardrails get in the way, and what they cost to run in a working day.
Before you start
- Available on Immersive One Advanced.
- No setup or configuration is needed. Each range gives you a scenario and a pre-built environment to work from.
- Because ranges are provisioned ahead of time, exercises are typically scheduled rather than started on demand.
What you'll get out of it
- A scenario your red team runs together, with a choice of AI provider.
- A live view (for facilitators) of how teams are progressing.
- A debrief at the end with high-level metrics: cost and token usage broken down per team and per model, and tasks completed. You can also export this data from Data Explorer.
How to set up and run an exercise
- In the navigation, open the Exercise dropdown and go to the featured section for Offensive AI Ranges.
- Browse the scenario library and open a scenario to see its details: the scenario itself, the infrastructure, the MITRE mappings, and the available AI providers.
- Select the scenario you want to run.
- Complete the configuration form: choose your teams, set the schedule, and pick your AI provider (one, or all of them).
- Submit. The platform handles scheduling automatically – it invites participants and stands up the infrastructure before the scheduled date and time.
- You can view your exercise on the dynamic threat range landing page.
- Click on the exercise to view it. Once the scheduled start time arrives, the environment auto-provisions and you can start the exercise.
- Join the exercise.
- During the exercise, use the facilitator view to watch live team scoring and usage, troubleshoot, and join the exercise to see how each team is doing.
- After the exercise, open the debrief page for the high-level metrics, including cost and token reporting per team.
What you'll see on screen
The key screens on each path, and what to look for on them.
Scenario page (facilitator). Before you run anything, opening a scenario shows its details in one place: the scenario, the infrastructure it stands up, the MITRE ATT&CK mappings it covers, and the AI providers available. Use this page to decide whether a scenario fits your team before you configure it.
Configuration form (facilitator). Where you set up a run: choose your teams, set the schedule, and pick your AI provider. Once you submit, the platform handles scheduling – you don't start the exercise manually.
Facilitator view (during the exercise). A view-only mode showing live team scoring, cost, and token usage as teams work. From here you can watch progress, troubleshoot, and join the exercise to see how each team is doing inside their range.
Offensive range (participant). Where you do the exercise, using the offensive tooling and AI agents in your range against the scenario infrastructure. You won't see whether individual answers are right or wrong here – that's by design.
Cost and token usage. Each team has a cost cap per exercise. As a team works, its token usage counts toward that cap. If the cap is reached, the AI pauses and the team continues in "human mode" for the rest of the range. Keep an eye on usage so you're not caught out.
Debrief page. Shown at the end. High-level metrics including cost and token usage broken down per team and per model, and tasks completed. This is where feedback lands. You can export the same data from Data Explorer.
Troubleshooting
| Issue | What to try |
|---|---|
| I can't find Offensive AI Ranges | Look under the Exercise dropdown in the navigation, in the featured section. |
| I can't set up a range | You need an AI Ranges license and the ranges administrator permission. Check these are in place. |
| The AI stopped responding partway through | Each team has a cost cap per exercise to prevent runaway costs. If you reach it, the AI pauses and you continue in "human mode" for the rest of the range. Watch your token usage – a facilitator can restart the range if needed. |
| My range didn't start | Occasionally a range fails to deploy. It will try to recover automatically. If it doesn't, contact support. |
Limitations
- Release one covers the offensive and developer personas only.
- Detailed after-action reporting for the AI parts is limited today, beyond the debrief page and Data Explorer export.
- You can't bring your own harness, AI, or tools yet – the range provides the available tooling.