MAD REFERENCE

What governed AI means for a physical venue

A concise reference for governed AI in physical venues: authority, decision fields, WorldModel responsibilities, safety precedence and implementation evidence.

Discuss your project Browse insights

DEFINITION

Governed AI connects an action to its authority.

In a physical venue, an AI-assisted system may select content, answer a question or propose a change. Governance defines the information and sources it may use, the actions it may take, the people who can intervene and the evidence needed to understand the result.

WorldModel™ provides a reference architecture for those responsibilities. The current WorldModel.global reference defines ten architectural layers and eleven cross-cutting policies. A project specifies and tests the functions it implements.

PRECEDENCE

Keep permission, safety and recording distinct.

CGL™ evaluates proposed action under the applicable authority and rules. OSOL™ has immediate hard priority when invoked. AAL™ is a non-gating, append-only observer: recording a decision does not approve or delay execution.

The ten layers are responsibilities, not ten sequential approval gates. Spatial governance operates through those responsibilities and the applicable policies. WorldModel OS supplies coordination interfaces; it is not an additional runtime layer.

TECHNICAL DETAIL

The ten responsibilities.

Open the engineering detailClose this sectionLook for the red circuit symbol for engineering detail.

Component	Name	Responsibility
VS+C™	Value System + Constitution	The venue's priorities, rights and protected limits define the purpose of its rules.
CGL™	Cognitive Governance Layer	Proposed actions are authorized, blocked or referred against the rules that apply at the time of action.
TGF™	Temporal Governance Framework	Schedules, time-limited permissions and shared-resource windows establish when an action is appropriate.
ICL™	Identity Continuity Layer	Language, access settings and relevant preferences continue within consent and purpose limits.
EDE™	Environmental Dynamics Engine	Current space, occupancy, conditions, content state and location-specific rules inform decisions.
MAOL™	Multi-Agent Orchestration Layer	Specialist agents receive bounded tasks and coordinate under the authority granted.
FCL™	Federation & Coordination Layer	Venues, operators and jurisdictions can coordinate while retaining local authority.
RGL™	Resilience & Graceful Degradation Layer	Reduced capability leads to a defined operating response, with protected requirements given priority.
OSOL™	Operational Safety Override	Designated safety actions have hard priority over experience, personalization and orchestration.
AAL™	Assurance, Analytics & Audit Layer	A non-gating, append-only observer records what was decided, by whom and under which rules and conditions.

TECHNICAL DETAIL

What the implementation brief should identify.

Open the engineering detailClose this section

Field	Question to resolve
Purpose and request	What result is requested, and which visitor or operating requirement does it serve?
Context and provenance	Which observations, content sources and versions support the proposed response?
Consent and identity	Which preferences may be used, for what purpose and for how long?
Authority and policy	Who or what may authorize the action, and which rules apply now?
Time and place	Which zone conditions, schedules and resource limits affect the decision?
Execution and fallback	Which interface can carry out the action, and what happens if it cannot?
Human and safety control	Who can pause, override or recover the system, and how is safety priority enforced?
Record and outcome	What was proposed, authorized, executed and observed, and which evidence is retained?

These are specification questions, not a published API schema. The implementation defines the actual fields, interfaces, permissions and tests.

IMPLEMENTATION

Use evidence at the level of the function.

Identify working functions, demonstrations, further development and roadmap items separately. Test the permitted path together with missing consent, unavailable services, rejected requests and recovery. The reference provides a design basis; commissioning establishes the behavior of the agreed installation.

WORK WITH MAD

Which decisions will your system be allowed to make?

Use this reference to frame the authority, interfaces and evidence your project needs, then discuss the implementation scope with Mad.

Discuss governed operation

Terms on this page

AAL™: Assurance, Analytics & Audit LayerThe WorldModel observer that records actions, decisions and their context. It does not approve, delay or block execution.
AI: Artificial intelligenceSoftware that can interpret information, generate content or propose actions. A venue still defines what it may do and what evidence is required.
API: Application programming interfaceA defined way for one software system to request information or actions from another.
CGL™: Cognitive Governance LayerThe WorldModel responsibility that authorizes, blocks or refers proposed actions against the rules and authority that apply at the time of action.
EDE™: Environmental Dynamics EngineThe WorldModel responsibility that supplies relevant space, occupancy, flow, environmental and content state for the place where an action is proposed.
FCL™: Federation & Coordination LayerThe WorldModel responsibility that supports coordination between venues and operators while preserving each participant's local authority.
ICL™: Identity Continuity LayerThe WorldModel responsibility that carries language, accessibility settings and relevant preferences within the visitor's consent and the permitted purpose.
MAOL™: Multi-Agent Orchestration LayerThe WorldModel responsibility that coordinates specialist software agents through bounded tasks under the authority they have been granted.
OS: Operating systemSoftware that manages the operation of other software and system resources. WorldModel OS is Mad's name for the interfaces connecting venue systems to governed coordination.
OSOL™: Operational Safety OverrideThe WorldModel responsibility that gives designated safety actions hard priority over experience, personalization and orchestration.
RGL™: Resilience & Graceful Degradation LayerThe WorldModel responsibility for a defined operating response when capability is reduced, while giving protected requirements priority.
TGF™: Temporal Governance FrameworkThe WorldModel responsibility for the time context of an action, including operating regimes, time-bounded permissions and mutually exclusive windows.
VS+C™: Value System + ConstitutionThe WorldModel responsibility that holds the venue's priorities, rights and protected limits, defining the purpose of its operating rules.