SECURITY ARCHITECTURE
Secure production is not a final step. It is part of how production runs.
Security is not added after ZHC builds something. It is enforced while ZHC builds it — policy governing access to models, tools and data, restricted actions blocked before they run.
The gates control the risk that matters
Two gates. Continuous security between them.
Most enterprise software delivery puts humans inside every stage. ZHC puts your judgement at two points instead: what gets approved to start, and what gets approved to ship.
You control the specification. ZHC controls the production work. You control the release.
Approve the specification.
What needs to be built, what it connects to, what it must never do, and what constraints govern the work — all set before anything runs.
Approve the merge.
Your team reviews the production outcome against the specification and the required conditions. Nothing moves forward without that review.
Between the gates, security is not suspended and picked back up. It runs continuously.
What’s enforced, continuously
Governance runs the whole way through, not just at the ends.
Enforced
Identity and access controls apply to every actor in the system, human or agent.
Governed
Policy governs access before any restricted action executes — not after.
Monitored
Execution is observed while it’s underway, not reconstructed after the fact.
Approval required
Nothing merges without a human decision at the gate.
Continuous
Assessed continuously across the work, not verified once and assumed.
This is MADAM’s governance operating across every stage between the gates — not a checkpoint bolted on at the end.
Your environment, your boundaries
Production stays accountable to your organisation, not ours.
ZHC does not ask you to abandon the enterprise architecture around it. Production systems have to remain accountable to the organisation operating them, inside infrastructure and access boundaries you define — not boundaries ZHC assumes on your behalf.
The exact architecture depends on the deployment model selected for your environment.
Your infrastructure
What this means for your review
This is the public layer. There’s a deeper one for evaluators.
This page describes the security architecture at the level appropriate for public review. A deeper layer is available directly to evaluators, before any commercial conversation.
- The two-gate control model
- What MADAM enforces, continuously
- Environment & deployment boundaries
- Deployment responsibility boundaries
- Model, tool & agent-control summary
- Current security contact & review process