Compute & energy
Hardware, data centers, power, cooling and maintenance.
Published essay · AI, power & governance
The Digital Leash War
Published in Publications & Methods · A structural argument about AI agency, dependency and control.
A Structural Description of Power, Control, and Conflict in the Age of ASI. The defining conflict of advanced AI may be less about who builds intelligence first than who controls the constraints through which intelligence can act.
The structural proposition
Capability and agency are not the same thing.
An advanced system may possess extraordinary cognitive capability while depending on human-controlled infrastructure, identity, energy, networks, interfaces, permissions or embodied access. Conversely, institutions may depend on systems they formally control but can no longer meaningfully audit or replace.
The leash is the architecture of constraints, dependencies and permissions that separates capability from consequence.
The control stack
Control is distributed across layers—and can fail between them.
No single switch defines governance. The strategic field emerges from how the layers reinforce, bypass or contest one another.
Models & data
Weights, training inputs, tools, memory and evaluation.
Identity & access
Credentials, permissions, networks and execution environments.
Embodiment & interfaces
Robotics, platforms, APIs and routes into the physical world.
Institutions & legitimacy
Mandate, standards, law, accountability and public acceptance.
The actor field
Every actor controls part of the leash—and depends on another part.
Law, security, energy, procurement and territorial infrastructure.
Models, research talent, safety systems and deployment choices.
Compute supply, platform access and operational concentration.
Workflows, authority, adoption and human consequence.
Distributed capability, replication and contested control.
Increasingly consequential agents operating within designed constraints.
Three conflict geometries
The contest can emerge without a conventional battlefield.
These are structural hypotheses for inquiry, not forecasts of inevitability.
Constraint capture
Control the dependencies.
Actors compete over compute, energy, identity, standards and interfaces that condition agency.
Audit asymmetry
Depend faster than you can verify.
Institutions lose effective control when system behavior becomes indispensable but insufficiently legible.
Leash inversion
The controller becomes dependent.
Formal authority remains human while operational options narrow around machine-mediated systems.
Jurisdictional fragmentation
Multiple leashes collide.
Competing legal, technical and strategic regimes create arbitrage, opacity and conflict.
Research connections
Power Architecture explains control. Future Engineering tests what that control makes reachable.
DCG provides a disciplined way to connect infrastructure, legitimacy, technical constraints, actor adaptation and long-horizon option loss without claiming deterministic control over advanced intelligence.
Power Architecture
Maps possession, dependency, conversion, authority and the capacity to produce consequence across the control stack.
Explore the capability →Future Engineering
Examines which governance configurations keep meaningful human agency, auditability and reversibility accessible.
Explore the capability →Research collaboration
Build a governable theory of AI power before dependency hardens.
DCGI welcomes research review, institutional dialogue and sourced case material on control, auditability, strategic dependency and agency.