FAQ

Does any of our data reach Compiled? No. Communications and agent actions are inspected inside your tenant. The only inbound flow is signed detector updates; the only outbound flow is findings to your own SOC. See Architecture.

What does the inspection AI use? Your own in-tenant Azure OpenAI resource by default (same model/vector space as the shipped detectors — no recompile), or a fully local model for air-gapped deployments. External OpenAI is never used unless you explicitly opt in, and the deployment fails closed rather than silently egressing.

How is it deployed? As an Azure Marketplace managed application into your own subscription. The runtime, storage, and audit logging run on your infrastructure. See Deploying Compiled.

Is it a blocking gateway or a triage layer? For communications, it's a near-real-time triage layer — it surfaces the elevated few for review; it does not hold a person's message. For AI-agent actions, it can decide inline (allow / flag / block) before the action runs.

How fast is the agent-action decision? The inline decision is designed for sub-100ms in-tenant. See Plugging Compiled into your agents.

What does a finding contain? Scored metadata only — timestamp, channel, sender, concept, severity, risk, policy attribution. Never raw content. See Findings & the dashboard.

How do you keep up with new threats? Detectors are refined continually and delivered as signed, version-pinned updates you stage and can roll back — with no access to your data.

Is there a SOC 2 report / security questionnaire / BAA? SOC 2 (in progress), a pre-filled SIG/CAIQ questionnaire, and a BAA are available under NDA — contact your Compiled representative.

What does it cost to capture every Teams message? Microsoft removed metering from the Teams change-notification APIs in 2025, so full-coverage capture carries no per-message charge; the only marginal cost is in-tenant compute, which is modest.