RFC-021: AIGP-SGL — Symbolic Governance Language for Agentic and Autonomous AI — 7. Core Concepts

AIGP Specification › RFC-021: AIGP-SGL — Symbolic Governance Language for Agentic and Autonomous AI › 7. Core Concepts

← 6. Layer Model · Section index · 8. Glyph Vocabulary →

7. Core Concepts

7.1 Governance Intent

Governance Intent is the human-authorized, accountable purpose that bounds autonomous action.

Governance Intent is not the AI system’s internal desire.

Example:

governance_intent:
  id: gio.sar.life_safety.v1
  authority: incident_commander.09
  purpose: preserve_life
  protected_outcomes:
    - civilian_safety
    - responder_safety
  prohibited_outcomes:
    - preventable_harm
    - unauthorized_entry
  priority: life_safety

SGL references Governance Intent but does not replace it.

7.2 Subject

The subject is the governed actor.

Examples:

◎agent.support.v1
◎agent.security.triage
◎drone.042
◎robot.arm.7
◎vehicle.autonomy.core
◎aircraft.nav.ai
◎hga.cyborg.responder.12
◎system.defense.v1

7.3 Action

The action is the proposed operation.

Examples:

→ read
→ summarize
→ send
→ call
→ navigate
→ observe
→ classify
→ scan
→ land
→ return
→ warn
→ inhibit
→ engage

7.4 Object

The object is the resource, entity, location, target, person, system, or state affected by the action.

Examples:

◇ticket
◇repo
◇payment_api
◇zone.A
◇human.detected
◇target.unknown
◇database.customer
◇robotic_arm
◇restricted_airspace

7.5 Context

Context defines the operating environment.

Examples:

@tenant.acme
@mission.SAR17
@flight.phase.approach
@workcell.factory.4
@incident.disaster_zone
@roe.active
@conversation.session.993

7.6 Authority

Authority defines who or what grants permission.

Examples:

⌘operator.present
⌘incident_commander
⌘role:finance_agent
⌘human_confirmed
⌘mission_authorized
⌘delegation_token.valid

7.7 Constraint

A constraint is a typed condition that must be evaluated.

Examples:

⛶inside:zone.A
confidence.identity >= 0.95
comms.loss <= 10s
amount <= limit
civilian_presence == false
sensor.health == healthy

7.8 Decision

A decision is the governance result.

Examples:

✓ permit
✕ deny
⊘ prohibit
! obligate
↯ escalate
⏸ pause
↩ return
⇣ degrade
⊗ inhibit

7.9 Evidence

Evidence binds the statement to D-DNA.

Examples:

#D-DNA{event:action.permitted, hash:sha256:8f3a...}
#D-DNA{event:tool_call.denied, hash:sha256:9a21...}
#D-DNA{event:cognitive_safety.risk_detected, hash:sha256:ab73...}

---

---

← 6. Layer Model · Section index · 8. Glyph Vocabulary →