AI AUTHORITY AND INSTITUTIONS
Public companies, algorithmic executives, autonomous businesses, synthetic state actors, municipal AI, fiduciary duty, contestability, and operational assurance—separated by function, law, evidence, and human authority.
“AI-run” can describe a company that sells AI, an underwriting model that determines unit economics, a board tool with veto influence, a multi-agent business, a synthetic political persona, or a city platform that coordinates infrastructure. The collection separates those claims before asking who still decides.
DIRECT ANSWER / EVIDENCE BOUNDARY
NEW ACCOUNTABILITY OBSERVATORY
Trace the principal, mandate, evidence, authority, external effect, and remedy behind each AI title or delegated agent.
Trace authority and remedyNEW LIFECYCLE AND REMEDY OBSERVATORY
Track material change, version control, correction propagation, incidents, and retirement after a system leaves the lab.
Trace material change and correctionPUBLIC-MARKET SNAPSHOT
The report-derived categories help distinguish an operating dependency from a product label, a thematic fund, or an executive-branding experiment.
Companies whose product, data layer, inference, or enterprise workflow is marketed as the central AI operating environment.
Caution: Product centrality does not prove customer outcomes, current growth, or autonomous governance.
Businesses where models materially shape lending, insurance, diagnostics, or other revenue-producing decisions.
Caution: Automation, fairness, regulatory status, and performance must be evaluated separately.
Funds that package robotics, semiconductors, data, software, or automation exposure under an AI theme.
Caution: Holdings and exposure change; the label does not make a fund a pure AI portfolio.
AI systems branded as executives, board observers, workflow directors, or organizational agents.
Caution: Public titles do not establish legal office, fiduciary capacity, or independent control.
NINE LEVELS OF PRACTICAL AUTHORITY
The same model can be low-risk at one level and institutionally decisive at another. Classification must follow the function actually delegated.
Retrieve or present information without ranking people, resources, or outcomes.
Can the source and age of each statement be inspected?Draft, summarize, translate, or calculate while a human retains the substantive decision.
Can the human independently inspect and revise the work?Generate one or more courses of action without automatically committing the institution.
Are alternatives, contrary evidence, and uncertainty visible?Rank cases, investments, petitions, inspections, or resources and thereby shape what receives attention.
What is hidden below the threshold, and who can change the ranking?Perform a bounded action that can be promptly undone, such as scheduling, routing, or drafting a provisional notice.
Is rollback tested, logged, and available without the same model?Spend money, alter access, issue a decision, discipline a worker, or change a resident’s practical rights.
Which independent human authority must approve, review, and correct?Sign, contract, certify, adjudicate, or otherwise create a legal obligation or public act.
Who has legal capacity and bears the nondelegable duty?Speak, negotiate, campaign, or appear as the organization, government, or officeholder.
Can the public identify the human principal and distinguish performance from authority?Direct strategy, assets, agents, and legal wrappers with little or no routine human operation.
Which natural or legal person remains responsible, solvent, reachable, and capable of shutdown?FUNCTION BEFORE TITLE
Product dependence, executive influence, business autonomy, state representation, and municipal administration have different legal principals and rights effects.
| Institutional form | Unit of analysis | Output | Formal authority | Critical boundary |
|---|---|---|---|---|
| AI-native public company | Revenue model, product, workflow, or underwriting engine | Software, decision support, automation, risk estimate, or service | Corporate management and board retain legal authority | AI dependence is not the same as AI legal control |
| Algorithmic executive | Strategy, workflow, investment, or operating decision | Recommendation, veto, coordination, or delegated action | Human directors, officers, and entity adopt or authorize | Title and practical influence do not create fiduciary personhood |
| Autonomous enterprise | Multi-agent task graph, tools, payments, and legal wrapper | Transactions, operations, contracts, or business continuity | Depends on external identity, policy, wallet, and entity controls | Operational autonomy does not prove lawful ownership or unlimited mandate |
| National proxy governance | Public service, administrative case, policy, or representation | Triage, recommendation, communication, allocation, or official act | Constitutional and statutory officeholders remain responsible | Synthetic presence does not create democratic legitimacy |
| Municipal cognitive system | Infrastructure, mobility, environment, service, or resident case | Forecast, dispatch, maintenance, permit, benefit, or alert | Municipality and authorized officials retain public-law duties | Coordination efficiency cannot replace notice, appeal, and public purpose |
EVIDENCE-QUALIFIED CASE REGISTER
The register records what the submitted reports use each example to illustrate—and what the public title or report does not prove.
16 cases shown
Ai Operating Platform
Enterprise Ai Vendor
Agentic Automation
Applied Ai
Clinical Data Platform
Board Advisory
Virtual Executive
Symbolic Executive
Board Observer
Agentic Enterprise Experiment
Government Ai Integration
Political Representation
Judicial Assistance
Urban Operating System
Regional Case Study
Platform Urbanism
No cases match the local filters.
NINE RESEARCH ROUTES
Each route carries source states, risks, controls, unresolved questions, and a retained-report trail.
Distinguish companies that sell AI, companies whose unit economics depend on AI, data and inference suppliers, thematic funds, and experiments in algorithmic management.
Open the research routeExamine AI board observers, advisory votes, virtual CEOs, workflow coordinators, and the humans who remain the legal directors, officers, owners, and fiduciaries.
Open the research routeFollow the shift from deterministic automation to agents that plan, call tools, coordinate sub-agents, move money, and operate through legal entities.
Open the research routeSeparate administrative decision support, synthetic ministers, political avatars, sovereign AI infrastructure, judicial triage, and actual legal authority.
Open the research routeExamine urban operating systems, digital twins, 311 services, predictive maintenance, traffic control, environmental monitoring, public safety, and automated administrative decisions.
Open the research routeMap who owes duties when AI recommends, negotiates, prices, signs, or controls workflows—and how automated optimization can create collusion or oversight failures.
Open the research routeTrace authority across policymakers, boards, executives, civil servants, vendors, models, operators, and affected people so the machine cannot become an accountability sink.
Open the research routeTranslate governance into notice, explanation, impact assessment, worker participation, appeal, correction, monitoring, and retirement.
Open the research routeExamine context loss, model drift, recursive delegation, cyber compromise, vendor lock-in, compute cost, conflicting agents, and the gap between a successful model call and a trustworthy institutional outcome.
Open the research routeTEN GOVERNANCE GUARDRAILS
These controls are cross-report synthesis, not a substitute for jurisdiction-specific law, collective bargaining, financial regulation, or public consultation.
Every binding function has a responsible officeholder, director, officer, or public body that cannot disclaim the outcome to the model.
Record what the system may inform, recommend, prioritize, execute, sign, spend, represent, and never do.
Preserve data lineage, model and policy version, recommendation, human adoption, execution, and correction.
Legal, financial, rights, and safety constraints are enforced outside the model’s learned objective and conversational context.
The human has evidence, time, authority, workload margin, and a tested mechanism to reject or stop the action.
Affected people can learn that AI contributed, obtain reasons, seek human review, correct data, appeal, and receive remedy.
Workers and representatives receive notice, training, bargaining, surveillance limits, and protection from automated discipline.
Agents have spend limits, allowlisted counterparties, dual control, settlement checks, and idempotent recovery.
Identity, tools, models, updates, and data are authenticated; uncertainty forces safe, reversible, or read-only behavior.
High-impact systems expire unless current evidence supports renewal; vendor exit, rollback, and retirement are tested.
METHOD AND EVIDENCE DISCIPLINE
A bilingual public-interest atlas for understanding when AI assists an institution, when it exercises practical discretion, and which humans and legal entities remain accountable.
WORKING VOCABULARY
An institution delegates practical decision, coordination, or representation functions to an AI while humans and legal entities remain formally responsible.
A software or embodied interface presented as an agent, official, executive, candidate, worker, or representative.
An AI system with material influence over strategy, allocation, workflow, or operations, regardless of whether it holds legal office.
A business whose operational agents can pursue goals, call tools, coordinate, transact, and recover with limited routine human input.
A multi-step process in which a model plans, delegates, uses tools, observes results, and revises its actions.
A governance principle assigning an identifiable human or institution the duty and capacity to supervise, explain, stop, and remedy AI-mediated action.
Presenting a human or political choice as an unavoidable technical output in order to diffuse responsibility.
A situation in which harm is produced by many designers, data sources, vendors, operators, and institutions but no actor can reconstruct or own the decision.
Use of algorithms to assist or produce decisions about benefits, permits, services, enforcement, or other public-law consequences.
A coordination layer integrating municipal data, sensors, workflows, and applications across departments.
A computational representation of urban infrastructure or processes used for monitoring, simulation, forecasting, or planning.
Competitive harm caused or reinforced when pricing or allocation systems learn or implement coordinated market behavior.
A policy and infrastructure strategy seeking national or institutional control over models, data, compute, language, and deployment.
The practical ability to understand, challenge, correct, appeal, and obtain remedy for an AI-mediated decision.
RETAINED REPORT LIBRARY
The original Markdown remains protected under /docs. Public pages expose a bounded bilingual synthesis and do not independently verify every claim embedded in a report.
Submitted market and enterprise report
Submitted public-markets and enterprise report. Company performance, valuations, ETF holdings, enforcement dates, and market statistics are a dated research snapshot and require fresh primary filings or official sources before being presented as current fact.
Submitted state-governance report
Submitted national-governance report. It combines documented public-administration use cases, theoretical proxy-governance analysis, and rapidly changing claims about synthetic political actors; each category remains visibly qualified.
Submitted corporate-governance report
Submitted corporate-governance report. Examples of AI executives, board observers, fiduciary duties, antitrust exposure, and regulatory duties are used as research leads and governance patterns rather than legal advice.
Submitted autonomous-enterprise report
Submitted autonomous-enterprise report. Legal-entity structures, agentic payments, zero-member company theories, and protocol claims are jurisdiction- and date-sensitive; the public synthesis distinguishes observed deployments from conceptual architectures.
Submitted municipal-governance report
Submitted municipal-governance report. It blends practical municipal operations, conceptual cognitive-city architectures, regional case studies, and legal analysis. Current deployments and local claims require official verification before publication as operational fact.