Skip to content
AIGP

RFC-021: AIGP-SGL — Symbolic Governance Language for Agentic and Autonomous AI — 2. Motivation

AIGP Specification › RFC-021: AIGP-SGL — Symbolic Governance Language for Agentic and Autonomous AI › 2. Motivation

← 1. Abstract · Section index · 3. Design Thesis →

2. Motivation

The first generation of AI governance focused on prompts, model calls, responses, data use, audit records, and compliance checks.

Agentic and autonomous AI systems introduce a different governance problem.

They do not merely respond.

They act.

An agentic AI system may:

read files
invoke APIs
send messages
modify records
call tools
query secrets
delegate tasks
operate infrastructure

An embodied autonomous system may:

navigate
observe
classify
move
land
return
manipulate
warn
coordinate
alter physical state

A force-capable system may support or participate in actions where lawful authority, human judgment, distinction, proportionality, necessity, and protected status matter.

For these systems, governance cannot remain static prose. It must become symbolic, executable, and replayable.

AIGP-SGL exists to express:

Who is acting?
What action is proposed?
What object is affected?
What context applies?
What authority is required?
What constraints are active?
Is the action permitted, denied, inhibited, modified, or escalated?
What evidence must be produced?

← 1. Abstract · Section index · 3. Design Thesis →