AuxControl
See the AI running through your organization
AuxControl connects to your workspace read-only, discovers every AI model, agent and shadow tool in use, scores the exposure the way an underwriter would, and keeps watching so nothing drifts unseen. It is the loss-control half of the policy: the part that lowers the risk rather than transferring it.
Discovery, including the shadow
It inventories the AI actually in use, sanctioned or not, maps each tool to the people using it, and surfaces the ones nobody told you about. Most organizations are surprised by this list.
Scored like an underwriter
A versioned scoring engine turns that surface into a single risk index, on the same basis your policy is underwritten. The number your broker sees and the number you see are the same number.
Watched continuously
Live telemetry and scheduled syncs re-check posture, raise alerts when an agent drifts or goes silent, and produce reports a board can read. Risk is managed between renewals, not only at them.
How it works
Connect your workspace
Your admin grants read-only access to Google Workspace or Microsoft 365. No software to deploy, no code changes, no agent to install.
We assess the surface
It inventories the AI models, agents and copilots in use, maps them to your people, and flags shadow AI.
Risk is scored
The scoring engine turns that surface into a collective risk index, versioned so a score can be explained later.
Monitored continuously
Detection rules run around the clock against live telemetry, raising alerts on drift and producing board-ready reports.
Live in an afternoon rather than a quarter. The first sync reads your directory and AI activity within minutes.
What it reads, and what it never touches
AuxControl reads through the providers' own admin APIs. It never writes to your tenant, never opens file contents, and can be disconnected at any time.
What it reads
- Directory headcount, which sizes the monitored population
- AI application and model usage, which finds sanctioned and shadow AI
- Admin audit events, which detect drift and silent agents
What it never touches
- File and message contents, which are never accessed
- Anything written back to your tenant, because access is strictly read-only
- Your environment, because there is nothing to install in it
Adversarial testing and accreditation
Monitoring tells you what an agent is doing. Testing tells you what it would do under pressure. AuxControl runs deployed agents through a certification range: live tool calls, adversarial users and poisoned documents, producing a risk profile across the dimensions of agentic liability rather than a pass or fail.
A range, not a checklist
Agents face simulated adversaries and hostile inputs in a live environment. What comes out is a measured profile of how the agent behaves when someone is actively trying to make it fail.
Bound to the deployment
An accreditation is hash-bound to the configuration that earned it: model, scaffold, tools, guardrails and prompt. Change the thing materially and the accreditation is void, because it no longer describes what is running.
Evidence, never a condition
Results inform underwriting and renewal terms. They are never a condition of cover, and accreditation is not insurance: it is evidence about a system, not a promise to pay.
How you get it
AuxControl is loss control attached to an Auxilium policy, not software sold on its own. Binding a policy provisions it.
Included at bind
Every AI Liability policyholder is provisioned AuxControl free when the policy binds. No separate purchase, and everything needed to monitor your covered exposure is in it.
Upgrade for the whole organization
Policyholders who want to govern beyond the exposure their policy covers can upgrade: full workspace discovery, compliance frameworks, governance policies, scheduled reports and the audit trail.
Priced on the population
The upgrade is priced on the population pulled from your directory rather than on seats, because the risk spans the whole workforce and not just the people who log in. Annual, all features included.
In development under this name. Findings feed underwriting and renewal, and enrollment is never a condition of coverage.