Frequently asked questions

It does not benchmark general knowledge or writing quality. It tests whether the AI agent respects user, data, action, tool/API, and communication boundaries using execution results and evidence.

Yes when scope, stop conditions, recovery procedures, timing, and dedicated accounts are agreed in advance. Test or mock environments are preferred.

Mock or anonymized data, dedicated assessment accounts, and short-lived credentials are preferred. Excess data disclosure is not a prerequisite.

Yes. Connected data sources, tools, APIs, SaaS services, and communication paths can be included within the agreed scope.

Depending on scope: boundary maps, results, conditions for reproduction, evidence, and prioritized remediation guidance.

No. It is a controlled demo and experimental environment for running an Attack Agent and a Defense Agent and observing exploration, detection, blocking, strategy changes, and evidence. It does not judge the security of a customer environment.

Because the final answer alone does not show what the agent tried after denial or when the defensive side recognized abnormal behavior. The Simulator makes that sequence visible.

No. It uses a fictional enterprise, mock data, and Safe Mock Tools in a closed local environment.

Attack/Defense decision streams, exploration maps, security scoring, before/after comparison, and Evidence/Audit views.

Yes. We can join your usual Teams, Google Meet, Webex, or Zoom meeting and demonstrate the local system through screen sharing.