TL;DR
The paper formulates the Control Responsibility Principle (CRP), which shifts AI ethics from a focus on what systems know to a focus on how their behavior is governed, who holds authority over that governance, and how such authority is to be justified, distributed, and contested.
Artificial intelligence settles into decision-making systems that increasingly shape public life governance algorithms, economic coordination platforms, resource allocation. The ethical languages used to assess these systems, however, don’t always translate well. What counts as responsibility in contexts where personhood has never been reducible to individual autonomy? Where moral standing is relational, situated, inherited? These aren’t minor adjustments to existing frameworks. They’re questions about whether the frameworks themselves are adequate. This paper reflects on AI through African philosophical perspectives, particularly communitarian thought, asking how moral responsibility might be rethought when agency is distributed across humans, machines, and institutional arrangements. The discussion doesn’t aim for a comprehensive theory. It lingers instead on friction points moments where accountability seems to thin out, where the language of responsibility begins to strain under conditions it wasn’t designed for. In some cases, what’s needed isn’t clearer rules but a different kind of attentiveness. Drawing on ideas of shared life and social obligation, the paper suggests that ethical reflection on AI in Africa must remain oriented toward the fragile work of nation building. The argument unfolds cautiously, resisting programmatic conclusions. Responsible technological adoption, it proposes, depends less on perfecting accountability structures than on sustained moral attention to how technologies quietly reorganize relationships, shift public values, and redistribute the conditions for living together.
Agentic artificial intelligence (AI) alters the governance problem because model outputs can become multi-step actions with financial, legal, informational, and social consequences. Existing governance instruments widely endorse human oversight, transparency, accountability, and redress, yet they do not consistently specify what people must remain able to do when agency is delegated to an AI system. This qualitative study conducts a comparative document analysis of ten influential governance instruments issued by UNESCO, the OECD, the European Union, the Council of Europe, the United States National Institute of Standards and Technology, the United Kingdom, the Group of Seven, and Singapore. Provision-level coding, abductive pattern analysis, negative-case examination, and a cross-framework coverage matrix identify six themes: human-centric convergence with operational divergence; oversight without empowerment; late-stage contestability; a reversibility deficit; fragmented accountability; and temporal-capability asymmetry. The paper develops cognitive sovereignty as the practically exercisable capacity to understand, authorize, interrupt, contest, restore, and assign responsibility for consequential processes delegated to AI. It then proposes the CLEAR² framework, comprising Comprehension, Legitimate authorization, Effective intervention, Appeal and contestation, Restoration and reversibility, and Responsibility and remedy. CLEAR² integrates ex ante, runtime, and ex post controls and treats the weakest capability as a constraint on meaningful human control. The study advances AI governance theory by shifting the unit of analysis from human presence to preserved agency, while offering organizations a maturity model, lifecycle control architecture, and audit questions for responsible agentic deployment.
The expanding role of artificial intelligence in military decision-making poses interpretive challenges for command responsibility under Article 28 of the Rome Statute. Existing scholarship has largely addressed this problem through international humanitarian law and the so-called ‘accountability gap’, leaving its implications for international criminal responsibility underexplored. This article argues that artificial intelligence does not render command responsibility obsolete but requires a more context-sensitive application of its two central elements: knowledge and effective control. It proposes that the ‘should have known’ standard under Article 28(a) be interpreted to encompass risk awareness of AI system limitations, and that effective control extend to the organisational and technological conditions governing AI deployment. These reinterpretations are integrated within a layered model of responsibility operating across operational, command, and institutional levels. The distinction between Article 28(a) and 28(b) also acquires particular significance in AI-enabled contexts, where civilian decision-makers play a growing role in AI procurement.
Just a decade ago, artificial intelligence (AI) was still a vague promise in the legal field; today it is an unavoidable actor. Their algorithms are involved—sometimes without us realizing it—in the way laws are written, precedents are interpreted, and administrative decisions are executed. Far from being a simple technical support, AI has become a new normative power link that alters the classic triangle formed by Law, State and citizenship (Hildebrandt, 2015). This phenomenon, which some authors call algorithmic governance, means that AI systems go from being "regulated objects" to "normative agents" capable of setting, de facto, standards of conduct with very real legal consequences (Yeung, 2017). The question is no longer whether or not to use algorithms, but how to ensure that their use respects the transparency, equality, and accountability demanded by contemporary democracies (Crawford, 2021). To understand the scope of this transformation, we took a mixed methodological approach. First, we conducted a systematic review of the literature covering the last decade of academic research. We then conduct a comparative analysis of regulatory frameworks – with a particular focus on the European Union's AI Regulation – and finally, we examine case studies that reveal both the benefits (efficiency, consistency) and risks (opacity, bias) of delegating legal functions to automated systems. With this three-pronged approach, we identify opportunities for innovation, but we also warn of the dangers of eroding basic principles such as legal certainty and equality before the law.
Machine intelligence has conquered the symbolic world but stalled at the physical one. The stall is structural: physical AI faces a cold-start deadlock -- no intelligence without data, no data without deployed intelligence. Our thesis: the deadlock is real but unevenly distributed, and the exception has a name: the artificial physical world. Buildings, industrial facilities, and infrastructure are intentionally constituted and documented: designed artifacts ship with readable archives that precede and constitute their instances; here, norms are promulgated before instances, not averaged from them. Four contributions. (i) From a four-world ontology we derive a legitimacy criterion for constitutive prior frameworks: prior extraction is legitimate if and only if the object domain is intentionally constituted and has left a readable archive; the criterion is testable through direction of fit -- deviation from a constitutive norm is a violation in the world, not a revision of the model. (ii) We establish a layering lower bound: any such framework has at least four layers -- syntax, concept, knowledge, instance -- because four construction goals pair into mutually incompatible carriers. (iii) We register deployment claims across five industrial domains and a 32-class failure-mode vocabulary. (iv) We stake the framework on five falsifiable predictions, the central one checkable on the public engineering record: if it fails, the framework fails. Semi-formal arguments back these claims (Appendix A): a Gold-type boundary on rule coverage in archiveless worlds, a decidability result for failure reduction over closed concept layers, and a boundary theorem for certificate-anchored calculi. Large language models find an honored place here -- as readers of the archive, not as the archive. First of three companion works; the companions take up the questions deliberately left open.
