05 / GOVERNANCE

Intelligence under legitimate authority.

The institution is designed to be autonomous within defined boundaries—not autonomous in defining those boundaries.

AI-NATIVE ARCHITECTUREEVIDENCE-LED RESEARCHGOVERNED INTELLIGENCE
GOVERNANCE MODEL

A hierarchy of authority, not a hierarchy of confidence.

Proposals, decisions and permissions are distinct. High-quality AI reasoning cannot confer powers that policy has not granted.

01

Chairman authority

Reserved decisions, capital ceilings, institutional amendments and immediate pause controls remain with the Owner.

02

Executive coordination

The proposed CEO function coordinates priorities, resources and exceptional cross-functional matters.

03

Independent challenge

Risk and protected assurance functions can challenge or block actions according to their mandates.

04

Deterministic controls

Software validates permissions, limits and institutional state before accepting consequential commands.

05

Recorded decisions

Material actions preserve their authority, evidence, dissent and rationale for later review.

06

Controlled evolution

Improving the institution does not permit it to rewrite its constitution or bypass its referee.

THE HUMAN BOUNDARY

Autonomy must be earned.

TCG’s first operating environment is designed for simulation. Real-capital authority requires separate commissioning evidence and explicit Chairman authorization.

The Owner can pause new exposure. Restarting after a pause or material failure requires appropriate revalidation; silence never expands authority.

OUR DESIGN PRINCIPLE

TCG can earn eligibility for live authority. It cannot grant live authority to itself.