From AI Ethics Principles to Implementation Commons

Methodologically, the paper is a reconstructive and normativeinstitutional-design argument. It does not offer empiricalvalidation of an existing association.Philosophically, the paper undertakes several distinctmoves. The public interest is recast as the structural multiplicityof social roles within a single subject, dissolving thechoice between an aggregation of stakeholder interests andan abstract defence of fundamental rights. The critical literatureon first-wave AI ethics is read as a set of design requirementsfor the institutional mechanism of the second stage;the body of critique thereby becomes material for designrather than evidence that the enterprise itself has failed. Thegeneral-purpose character of AI entails a meta-institutionalrequirement: the design object becomes a mechanism forgenerating institutional layers across contexts, since no singlelayer can encompass the open-ended set of deploymentcontexts. The philosophical question of the relation betweenethics and rationality yields three operational requirements:a rationally articulable link between ethical principles andrisk, harm, rights and the evidence base; the academic freedomof the community as a condition of its self-correction;and an education for practitioners that combines technicalcompetence with normative formation.Conceptually, the paper introduces a number of distinctionsnew to the AI governance literature. Normative traceis defined as a structural property of a technical artefact thatmakes it an accountable implementation artefact and preservesthe link between the technical layer and the values itwas built to protect. The implementation commons is introducedas a category in its own right, set apart from a technologyfoundation by the presence of a normative trace in everyartefact, and from a standards-setting body by the presenceof technical implementation. A dedicated representationaltrack for deployers within governance bodies is a structuralcondition of multistakeholder balance, since without it suchan association drifts in practice towards provider dominance.A two-condition framework for staged delegation defines theoperational criterion for transferring authority to AI: maturityof the human loop and institutional readiness of the recipientlayer. The architectural bottleneck of the AI transitionis derived from the convergence of three independenttheoretical foundations: the limits of human cognitive bandwidth,the law of requisite variety, and the logic of autonomiccomputing. This makes aggregation a structural ratherthan sectoral requirement.By way of synthesis, the paper draws together bodies ofliterature not previously combined within a single framework.The logic of complementarity in general-purposetechnologies and the paradigm of smart regulation are linkedthrough a reading of first-wave AI ethics as the mechanismby which criteria for assessing benefits and harms areformed for contextual regulation. The economic logic of ageneral-purpose technology and the political logic of multistakeholderrepresentation converge on the same institutionalrequirement, reducing the proposal’s vulnerability to ashift in either logic. Two independent analytical frames, onenormative and one architectural, converge upon the same institutionalform; this is not in itself a proof of its correctness,but it lowers the probability that the proposal holdsonly within a single analytical perspective.Historically and strategically, the paper takes several stepsthat go beyond the conventional discourse on principles inAI ethics. First-wave AI ethics is read as a structural breakthrough,in that it furnished the rational basis on which thefirst comprehensive horizontal regulatory framework for AIwas built, irrespective of whether this was its historical intention.AI ethics, and the regulation that rests upon it, arerecast as a mechanism that exposes longstanding organisationalpathologies which the AI transition renders ruinous,rather than as a source of new costs for organisations actingin good faith. The AI ethics community is treated as a strategicsubject facing an institutional choice with operationalconsequences; declining the initiative leads to the community’sdisplacement by other actors within the emerging infrastructureof the second stage. The role of AI ethics asmethodological steward is paired with explicitly embeddedmechanisms of self-scrutiny, which close the regress of accountabilityand keep the community open to external contestation.The anthropomorphisation of AI is treated as astructurally embedded risk, engendered by the very terminologyand architecture of the technology, which transferspart of the responsibility from the user to the design of theinteraction interface and to regulatory requirements for AIliteracy. Version v1.1 includes Supplementary Material A and places the shared reference list after the main text and supplement. Copyright © Dmitrii Krivosheev, 2026. This work establishes and publicly timestamps the concepts of: Implementation Commons Architectural Readiness architectural mediation layer staged governance integration protocol