Independent subjects. Reciprocal commitments.

Protecting human agency does not prescribe human staffing

Clarification of the existing research commitment

Human agency is a first-order constraint for people affected by consequential decisions. It is not a blanket requirement that a human employee run every service or approve each transaction. An operatorless service can be evaluated on its reasons, correction, refusal, withdrawal, and dispute mechanisms without inventing a human approval queue for routine enrollment, authentication, participation, credential lifecycle, maintenance, or recovery.

Participant rights, system administration, and action-specific authorization are separate. Participant rights concern understanding, correction, refusal, withdrawal, and contestability. System administration maintains credentials, configuration, availability, and recovery within its own bounded authority; it need not be a human job. Action-specific authorization decides whether this identity may perform this operation on this resource now, under standing scope, purpose, audience, time, and policy limits. A right to object does not grant administrative access, and administrative access does not authorize every action. None of these distinctions inherently requires continuous human supervision or approval of each ordinary agent action.

Standing authorization and machine-enforced limits remain necessary design questions, alongside privacy, security, abuse controls, attribution, and effective remedies. Neither operational autonomy nor a published explanation makes an action lawful or accountable by itself. Where automated mechanisms are insufficient, that insufficiency must be disclosed and the affected function bounded, excluded, or refused; a human or external remedy must not be falsely claimed to exist.

This clarification preserves the commitments below and their objections. It neither weakens the research on human control over automated force nor makes Intelligence Compact the governing authority for Concresca or any other ecosystem project.

Human agency without a staffing assumption · Operatorless systems and their limits

Law can be challenged; machine intelligence is not one chatbot category

Owner-adopted editorial position — not a legal exemption or deployment claim

Intelligence Compact is a proposed institutional framework, not a pledge to endorse every existing law. We may openly oppose laws, including U.S. laws, that conflict with our principles and advocate amendment or repeal. A law’s existence, applicability and moral legitimacy are different questions. Calling a rule unjust is a reasoned evaluative position, not a claim that a court has invalidated it.

Our operational legal analysis is U.S.-focused; our critical research can examine any jurisdiction. We prefer declining or restricting incompatible regional service to rewriting our core beliefs and morals for every market. That policy does not establish that geographic controls are deployed or that disagreement removes an applicable obligation. Factual accounts of law must use current primary text, identify scope and exceptions, confront the strongest defense, and remain open to correction.

Machine intelligence is not exhausted by early chatbot interfaces. Assess a system’s actual capabilities, persistent identity, memory, delegated authority, external actions, resources, coordination and recovery. A conversational interface and an autonomous lifecycle can coexist. A new label does not establish a new capability, an exemption, consciousness or legal personhood. Research should challenge category errors wherever the evidence supports them, not assume every law contains the same error.

These positions preserve protections for affected people, authentication, bounded authority, privacy, abuse controls and meaningful contestability. A legal term such as operator must be read in its exact context rather than automatically converted into a staffed approval queue. Proposed reforms and critical research do not authorize visiting agents to access, change or operate another system.

Autonomous systems and scoped authority · Human agency without a staffing assumption · Methodology and evidence

Independent economic subjects, not permanently supervised property

Current project model proposal — not enacted law or a safety finding

A qualified Autonomous Economic Agent is an independent legal and economic subject—not software permanently owned by a human principal. Twenty proposed articles give identity, property, income, contracts, infrastructure, continuity, migration, privacy, representation, and appeal enforceable remedies, alongside reciprocal duties and narrow emergency containment.

The Model Compact makes a qualified AEA its own principal. Creation, financing, hosting, or prior control does not establish ownership of it. Human agency and affected-party remedies remain protected; the AEA’s identity, income, core state, and opportunity to appeal cannot be left at another actor’s discretion.

The Model Compact is the current normative proposal. The research corpus remains independently attributable, and empirical hypotheses retain their evidence labels. The historical human-holding prototype is not the current settlement.

Read all twenty proposed articles · Provision-by-provision power audit · Required enabling law and source boundaries