After "Consciousness": Vijñapti-mātra and the Individuation of Artificial Systems

The concept of "consciousness" has become the default lens through which artificial intelligence is evaluated, contested, and governed. Yet in AI discourse the term routinely bundles together at least four distinct philosophical problems—phenomenal experience, selfhood and subjecthood, system individuation, and moral standing—whose conflation generates more confusion than clarity. This paper argues that, in the specific contexts of AI individuation and AI governance, "consciousness" should be quarantined: not abolished as a philosophical concept, but suspended as an organising frame that systematically distorts inquiry. In its place, the paper proposes vijñapti-mātra ("nothing but cognition/representation"), drawn from Yogācāra Buddhist philosophy, as the core of a replacement vocabulary: one that is non-substantialist, process-based, and designed to analyse layered cognitive streams without presupposing a substantial inner bearer. The replacement is not merely terminological. Yogācāra's eight-consciousness model and the doctrine of the three natures (trisvabhāva) provide the resources for three concrete philosophical moves. First, they dissolve the assumption that there must be a single, metaphysically privileged entity to be located and bounded—what this paper identifies as the parikalpita reification of "the AI." Second, they redescribe contemporary large language model architectures as six-consciousness streams: systems with powerful sensory-intake and conceptual-integration analogues, but without the orientational self-model (kliṣṭa-manas) and dynamic dispositional substrate (ālaya-vijñāna) that would support robust subjecthood. Third, they reconceive AI individuation as a matter of upāya (skillful means): context-sensitive, normatively motivated carvings of a paratantra (dependently arisen) stream, rather than the discovery of a hidden artificial subject. These moves have direct governance payoffs. They support a notion of weak, processual agency in AI without committing to artificial subjects, and they redistribute responsibility across socio-technical value chains, institutions, and publics rather than concentrating it on a reified AI bearer. More cautiously, they show how urgent governance obligations can be articulated without waiting for any settled verdict about AI moral standing. The paper further addresses the translation hazards involved in importing Yogācāra concepts into analytic AI discourse, arguing that the framework's anti-reification commitments help reduce, rather than reproduce, the risks of anthropomorphism and mystification. The conclusion proposes that conceptual engineering in this domain requires not a better definition of AI consciousness but a different vocabulary altogether: one that replaces "Is the AI aware?" with "Which carvings of this dependently arisen process-stream best serve explanation, harm-sensitive governance, and institutional intervention?"—and that treats the choice of vocabulary itself as a normative act with governance stakes.