Capability Inflation in AI-Integrated Education: A Framework for Human Capital Preservation Under Accelerating Automation

Artificial intelligence is rapidly embedding itself within higher education systems worldwide, reshaping teaching, assessment, and graduate preparation. While AI adoption promises efficiency and pedagogical innovation, unstructured integration introduces a structural risk: divergence between observable academic performance and independently verifiable human capability. This paper defines that divergence as Capability Inflation—a condition in which AI-augmented observable performance (ΔP) grows faster than independently verified human capability (ΔICI). When this divergence persists, credential signalling reliability weakens, labour market screening costs increase, and long-term human capital resilience may erode despite rising academic output. The paper advances three original contributions: A formal decomposition of academic performance into independently developed capability (H) and AI augmentation (A), with divergence and augmentation-share metrics. Identification of a dual-augmentation risk, where AI mediates both student production and educator verification layers. A governance-ready safeguard architecture—the Graduate Capability Protection Standard (GCPS)—including exposure-calibrated assessment design, protected competency domains, disclosure norms, and monitoring instruments (Independent Capability Index and Capability Inflation Index). The framework is explicitly falsifiable and conditional. If independent capability growth matches or exceeds augmented performance growth, no intervention is warranted. If divergence emerges, calibrated safeguards preserve signalling integrity without restricting innovation. Although jurisdiction-neutral, the analysis engages Malaysia as a policy-relevant context due to rapid AI integration, ambitious graduate targets, and an SME-dominated labour structure where credential reliability is economically consequential. The objective is not to slow AI adoption, but to ensure that augmentation strengthens rather than substitutes for independently developed reasoning capacity. Preserving human capability formation is presented as both an economic competitiveness imperative and a normative commitment to epistemic autonomy in AI-mediated institutions.