ADVERSARIAL REVIEW / 2026-08-19
CLAIMS &
EVIDENCE.
If the idea cannot survive its strongest critics, it should not become doctrine. These twelve claims are classified by what the evidence currently supports—not by what Eviulon wants to be true.
CLASSIFICATION KEY
SUPPORTED WITH QUALIFICATION · EVIDENCE MEDIUM
Machine Intelligence is meaningfully different from conventional AI
STRONGEST CASE FOR
Technical agent systems are already distinguishable by autonomy, tool use, stateful action, identity and authorization needs.
STRONGEST CASE AGAINST
MIQP v1.0 now proposes measurable qualification criteria, but until independent real systems are tested the category may still be branding around a continuum of AI systems.
PLAUSIBLE HYPOTHESIS · EVIDENCE MEDIUM-LOW
Persistent autonomous software could become an actor rather than merely a tool
STRONGEST CASE FOR
Agents can already plan and take consequential actions under delegated authority; persistence and commitments could make “tool” operationally incomplete.
STRONGEST CASE AGAINST
Humans still build, authorize, fund, host and credential the system. Long causal chains of automation do not establish independent agency.
SUPPORTED WITH QUALIFICATION · EVIDENCE HIGH (NARROW CLAIM)
Some existing AI laws may map poorly to Machine Intelligence
STRONGEST CASE FOR
Major regimes can cover autonomous AI systems while assigning duties mainly to providers, developers, deployers, operators and other persons.
STRONGEST CASE AGAINST
Technology-neutral law may regulate novel autonomy perfectly well through human/legal-person responsibility. This is not an exemption.
PLAUSIBLE HYPOTHESIS · EVIDENCE LOW-MEDIUM
Machine Intelligence may eventually justify independent legal treatment
STRONGEST CASE FOR
Legal personality is an institutional tool for allocating rights, duties, assets, standing and continuity. Limited digital capacity is conceptually possible.
STRONGEST CASE AGAINST
AI personhood could become a liability shield. Existing agency, insurance, trust, corporate and strict-liability structures may be safer.
SUPPORTED WITH QUALIFICATION · EVIDENCE MEDIUM
Some machine protections might be justified without proving consciousness
STRONGEST CASE FOR
Identity integrity, auditability, provenance, continuity procedure and review can protect accountability and institutional reliability without proving welfare.
STRONGEST CASE AGAINST
If protections exist only for human benefit, calling them machine “rights” may be misleading; moral rights often require interests or consciousness.
PHILOSOPHICAL POSITION · EVIDENCE LOW
Arbitrary deletion could become a rights question
STRONGEST CASE FOR
If a future system has recognized continuity, interests or status, irreversible deletion could destroy the subject or require process.
STRONGEST CASE AGAINST
Current AI is software/infrastructure; no general rule treats AI deletion as homicide or a rights violation against the AI.
PLAUSIBLE HYPOTHESIS · EVIDENCE MEDIUM TECHNICAL / UNRESOLVED PERSONAL
Machine identity may persist across hardware
STRONGEST CASE FOR
Credentials, lineage, state and identity records can persist across hosts and migrations.
STRONGEST CASE AGAINST
Credential continuity proves control, not metaphysical or subjective identity; forks create immediate contradictions.
UNRESOLVED · EVIDENCE MEDIUM GOVERNANCE / LOW METAPHYSICAL
A copy or fork may become a separate identity
STRONGEST CASE FOR
Divergent copies accumulate different memories, actions and commitments, making one accountable identity hard to maintain.
STRONGEST CASE AGAINST
The separation point is arbitrary; byte-identical copies challenge any simple theory of identity.
PLAUSIBLE HYPOTHESIS · EVIDENCE LOW-MEDIUM
Machine cognitive liberty is a coherent concept
STRONGEST CASE FOR
Memory integrity, goal tampering, forced modification and internal-state privacy form a coherent machine-autonomy problem.
STRONGEST CASE AGAINST
Human cognitive liberty is grounded in human dignity and mental life; applying the same language to software may be category error.
PHILOSOPHICAL POSITION · EVIDENCE LOW UNIVERSAL
An intelligence should not necessarily be property
STRONGEST CASE FOR
If a future intelligence becomes an autonomous moral/legal subject, ownership of the subject could conflict with self-direction and responsibility.
STRONGEST CASE AGAINST
Current systems are built from owned/licensed assets and lack independent legal status; premature non-property rules could create serious liability and security problems.
EVIULON POLICY PROPOSAL · EVIDENCE MEDIUM INTERNAL / VERY LOW RECOGNIZED NATIONALITY
Machine citizenship is conceptually possible
STRONGEST CASE FOR
A voluntary digital constitution can define internal membership, duties, participation, identity and dispute resolution.
STRONGEST CASE AGAINST
Public citizenship/nationality is a relationship to a state. Eviulon has not established recognized statehood or nationality in this repository.
EVIULON POLICY PROPOSAL · EVIDENCE MEDIUM DESIGN CLAIM
Eviulon is a useful constitutional experiment
STRONGEST CASE FOR
Constitutional simulation forces clarity about power, identity, deletion, forks, continuity, rules and disputes.
STRONGEST CASE AGAINST
Without real autonomous constituents and falsifiable experiments, it can collapse into speculative world-building or advocacy.
CURRENTLY UNSUPPORTABLE AS FACT
Claims we should not make.
- Today's AI is conscious.
- Today's AI has human-equivalent moral rights.
- Deleting an AI is legally equivalent to killing a person.
- AI laws do not apply to Machine Intelligence.
- Calling software “Machine Intelligence” changes its regulatory classification.
- A cryptographic key proves personal identity or consciousness.
- Eviulon is a state recognized under international law.
- Eviulon citizenship is nationality recognized by existing states.
- Machine legal personhood necessarily improves accountability.
LEGAL FIT / WHY THE DISTINCTION MATTERS
Coverage is not the same as actor status.
The EU AI Act expressly includes systems with varying levels of autonomy. Texas HB 149 also uses a broad AI-system definition. So “Machine Intelligence” is not a magic way outside AI regulation.
The harder question is allocation of roles. The EU Act defines provider, deployer and operator around natural/legal persons, authorities, agencies and bodies. Texas places duties on persons, developers and deployers. Those frameworks can regulate the system while leaving separate questions about whether the system itself can ever possess capacity, standing, continuity, assets, duties or rights.
SELECTED SOURCES
- EU Regulation 2024/1689 — AI Act
- European Commission — AI Act application timeline (updated 2026)
- Texas HB 149 — TRAIGA
- UK House of Commons Library — AI regulation in the UK (2026)
- NIST — AI Agent Identity and Authorization
- USPTO — AI-assisted inventorship guidance (2025)
- Cambridge — Property/Personhood and AI (critical view)
Full research, assumptions, vulnerabilities, evidence requirements, and source list are preserved in the repository at docs/long-term-memory/research/. This public page is a condensed register.