Guardrails, Compliance & Operator Behavior
Define scope, no-go zones, change policy, maintenance windows, compliance overlays and operator behavior that constrain and shape STRATUM AI execution.
Authority is part of the graph
AI Studio treats authority and policy as first-class DAG inputs. A sophisticated model is still not allowed to decide its own boundaries. Scope, No-Go Zones, change rules, maintenance windows and compliance overlays are wired into the same Operator that produces the plan.
The important distinction is enforcement: guidance can influence reasoning, but anything STRATUM can enforce should become a backend policy/authority rule so a bad plan is rejected even if the model proposed it.
Guardrail nodes
Guardrails
- Scope & Authority — Defines where the Operator may observe or modify resources and can separate broad visibility from narrower mutation authority.
- No-Go Zones — Defines explicit resources, datacenters, networks, fabrics or tagged systems that the Operator must not touch. These are hard boundaries, not suggestions.
- Change Policy — Adds rules governing what kinds of changes are allowed, what needs additional approval and what operating constraints must be satisfied.
- Maintenance Window — Constrains mutations to approved operating windows where change timing itself is part of policy.
Compliance overlays add enforceable checks and evidence
Compliance
- NIST Policy Pack — Activates STRATUM-enforceable checks/evidence mappings associated with the selected NIST-oriented policy profile. It does not by itself certify an environment.
- SOC 2 Policy Pack — Adds SOC 2-oriented operational controls and evidence expectations that STRATUM can evaluate or record. It is a policy/evidence overlay, not an attestation.
- Organization Policy — Adds site-specific rules that belong to the organization rather than a public framework—approved networks, artifact rules, encryption expectations, change controls and other local requirements.
A policy pack does not magically make an environment compliant. What AI Studio can do is make the STRATUM-enforceable controls, checks, approvals and evidence mappings visible and repeatable inside the Operator DAG.
Operator Behavior
Operator Behavior controls the character of the Operator: autonomy level, what to do when uncertain, whether to prefer existing resources, Guest Tools or approved Arsenal assets, whether to avoid downtime, whether to repair before replace, whether to snapshot before risky changes, and how much parallel work or explanation detail is appropriate.
Some behavior settings influence AI reasoning. Others map into Mission execution limits. AI Studio keeps both visible in one Operator instead of scattering them across hidden prompts and backend defaults.
Make “do not touch” visible and explicit
No-Go Zones are the clearest example of why the DAG is more than presentation. Drag the node onto the canvas, connect it into the authority/policy path, and configure the protected scope. The restriction becomes part of the Operator’s versioned graph.
When STRATUM can enforce a boundary, make it an authority or policy rule. The Operator may reason about policy, but the platform remains responsible for rejecting forbidden work.