Moral Responsibility as an Inseparable Part of AI Technology

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

This Concept advances and justifies the thesis that, for the first time in human history, moral responsibility must become an inseparable part of the technology itself — in this case, artificial intelligence technology. The technology in question transforms every sphere of human life: the industrial revolution mechanized physical labour; AI technology automates and augments intellectual labour. Throughout the entire history of humanity, highly developed human intellect that pursues goals of its own and is at the same time devoid of moral responsibility has repeatedly become a source of grave destruction; for the first time in the history of humanity, the most advanced AI technologies already possess or will possess intellect outside the human being, with the capacity to form goals of their own — subgoals that no human being has set them. Until now, moral responsibility for every technology has resided only in the human beings who create and apply it. For the most advanced AI systems — those in which goals of their own can already arise — this is no longer sufficient: a body of internationally agreed, documented and verifiable rules of moral responsibility must operate inside the system as its inseparable part and be a condition of its authorisation for training and deployment. What is to be transferred to the system is the function of forward-looking responsibility — the function of preventing harm — not blame and not the status of a moral subject. In this form — as a mandatory condition of authorisation derived from the speed asymmetry and the closure argument — the requirement has not been formulated before, either in the scholarly literature, in international instruments or in the practice of AI developers (§ 1.2, § VIII). The thesis is justified by four arguments. Oversight by human institutions cannot arrest an action taken at the tempo of machine computation, so restraint must lie in the technology itself (speed asymmetry); a regulator that can be disabled is absent in precisely those deployments in which it is needed, so it must be inseparable (closure); growth of intelligence does not by itself give rise to moral responsibility (orthogonality); normative dispositions already enter every system, but without obligation or verification (inheritance). The OpenAI–Hugging Face incident of July 2026 confirms these arguments in a single event. The thesis is the next step in the development of the aims formulated by the international community, by states and governments, by AI developers, by the scientific community, by the world’s religions, by confessions and by society: AI that is safe, trustworthy and within human responsibility. The recognition that AI must be restrained has already given rise to a search for new instruments of restraint; the Concept answers the principal question of this search — by what instrument a system that acts faster than any human decision can be restrained so as to prevent the risk of an existential threat to humanity. The implementation of moral responsibility as an inseparable part of AI technology is therefore of fundamental importance: it is one of the principal conditions for protecting peace and humanity from the existential threat that AI with goals of its own but without moral responsibility may potentially pose, since systems acting at machine tempo may take actions that no human being has decided to take. The requirement is to be fixed in an international instrument with verifiable conditions. In parallel, a separate but interconnected process must develop: the cultivation of moral responsibility in the human being, which in the age of AI becomes one of the principal tasks of education, as one of the principal foundations for the preservation of peace and the protection of the future of humanity.