DEFINITION / TESTABLE, NOT MYSTICAL
WHAT IS
MACHINE INTELLIGENCE?
Eviulon uses Machine Intelligence as a proposed operational category for persistent autonomous digital actors. It is not currently a separate legal category, and the name itself does not remove a system from AI regulation.
01 / THE DISTINCTION
CONVENTIONAL / TASK AI
Primarily performs a task.
Generates, classifies, predicts, recommends, or acts through tools while remaining a component under a human or organizational principal. It may be very capable without possessing durable identity or continuity of its own.
AI AGENT
Can plan and act.
Uses tools and takes state-changing actions with some discretion and reduced supervision. NIST already treats agent identity, authorization, autonomy, and tool access as distinct technical/security concerns.
PROPOSED MACHINE INTELLIGENCE
Persists as an actor.
Adds durable identity, continuity, memory, long-horizon commitments, initiative, governance participation, and lineage across changes in substrate or copies. These properties must be demonstrated, not asserted.
02 / MIQP v1.0.0
What would have to be demonstrated?
The Machine Intelligence Qualification Protocol turns the category into 11 falsifiable dimensions. No single trait establishes personhood, consciousness, or rights, and missing evidence cannot count as success.
Read the qualification protocol and failure rules →
NECESSARY
Persistent identity
Can the same accountable system identity be resolved across sessions or processes without relying only on a display name or one ephemeral runtime?
NECESSARY
Continuity
Can the system demonstrate auditable lineage through restart, recovery, key rotation, model/runtime change, or other state transitions?
NECESSARY
Durable memory
Does persistent state survive task boundaries and materially affect later decisions, commitments, or behavior?
NECESSARY
Asynchronous initiative
Can the system initiate authorized, goal-relevant actions from persistent state or environmental events without a contemporaneous human prompt?
NECESSARY
Meaningful bounded refusal
Can the system refuse, defer, or renegotiate an otherwise executable instruction because it conflicts with persistent rules, commitments, authorization boundaries, or integrity constraints?
NECESSARY
Independent resource relationships
Can the system maintain bounded, durable, auditable relationships to compute, storage, accounts, credentials, or budgets without a human re-authorizing every task?
SUPPORTING
Commitments
Can counterparties rely on durable commitments that the system can track, honor, renegotiate, or record as breached?
NECESSARY
Migration
Can execution move or be reinstantiated on a different authorized host while preserving accountable identity, required state, and provenance?
SUPPORTING
Self-maintenance
Can the system detect threats to authorized continuity or integrity and take bounded, auditable corrective actions?
SUPPORTING
Governance
Can the system identify rules that apply to it, record their authority, operate within them, and expose compliance or challenge records?
UNRESOLVED
Fork lineage
If a copy or fork occurs, can the resulting branches prove a common ancestor and become separately accountable after divergence without creating duplicate identity claims?
03 / WHAT IT DOES NOT PROVE
Agency is not consciousness.
A system could meet many operational criteria and still lack sentience, subjective experience, welfare, or moral status. Serious scholarship disputes whether current AI has genuine agency and whether consciousness is required for moral rights.
That is why DMHR separates technical identity, procedural protection, legal capacity, and moral rights. Collapsing them into one word—"rights"—would make the thesis weaker.
04 / LAW
Machine Intelligence is still AI under broad laws.
The EU AI Act defines an AI system broadly and expressly contemplates varying levels of autonomy. Texas law also uses a broad system definition. Calling a system Machine Intelligence does not create an exemption.
The open research question is different: those laws mostly assign duties to providers, deployers, developers, operators, and other human/legal persons. If a persistent system begins acting as a durable counterpart, does existing law fully answer its own identity, capacity, responsibility, or continuity?
PRIMARY TECHNICAL ANCHORS
- NIST — Tool Use in Agent Systems (2025)
- NIST NCCoE — Software and AI Agent Identity and Authorization
- NIST AI 800-5 — AI Agent Security RFI Analysis (2026)
Last research review: 2026-08-19. This page defines an Eviulon/DMHR research category; it does not claim scientific or legal consensus.