Student Adoption Intention, Academic Integrity Risk, and Learning Outcome Perceptions Under the NEP 2020 Framework

The emergence of generative artificial intelligence tools — most visibly ChatGPT, but including Google Gemini, Microsoft Copilot, and Anthropic’s Claude — has introduced what many educational scholars are describing as the most significant pedagogical disruption since the mass adoption of the internet. Unlike prior educational technologies, which augmented existing learning processes, generative AI can produce complete essays, laboratory reports, mathematical derivations, code, and examination answers indistinguishable from student-generated work, creating a fundamental challenge to assessment-based credentialing systems that have constituted the organisational core of higher education for over a century.In the Indian higher education context, this challenge arrives simultaneously with the implementation of the National Education Policy 2020, which mandates a shift toward competency-based, multidisciplinary, and experiential learning frameworks explicitly designed to move away from rote memorisation and examination performance as the primary educational currency. This creates a paradoxical institutional moment: NEP 2020’s ambition to reorient Indian education toward deeper learning and critical thinking is potentially aligned with the use of generative AI as a learning scaffold — if used to support exploration, research formulation, and iterative thinking — but is simultaneously undermined by AI’s capacity to circumvent precisely the assessment mechanisms through which NEP 2020’s competency gains are meant to be demonstrated and certified.The current study addresses this tension empirically through a Technology Acceptance Model extended with academic integrity risk and NEP 2020 policy alignment as theoretically motivated constructs, applied to survey data from 1,521 undergraduate and postgraduate students across nine Indian universities spanning Tamil Nadu, Andhra Pradesh, Maharashtra, Uttar Pradesh, and West Bengal. The study’s dual-outcome design — treating both learning outcome improvement and academic integrity violation risk as simultaneous consequences of generative AI adoption intention — reflects the field’s emerging consensus that these outcomes are not alternatives but co-occurring consequences of the same adoption behaviour, with their relative magnitude determined by pedagogical context, institutional policy clarity, and individual use orientation.The study’s Indian focus is motivated by three features of the Indian higher education context that are not captured in the predominantly Western empirical literature: the extreme heterogeneity of institutional policy responses, ranging from total prohibition in some state universities to active encouraged integration in IITs and private deemed universities; the language access dimension, where AI tools’ English language bias creates differential access by student linguistic background; and the high-stakes examination culture, where the pressure to perform in terminal written examinations creates a different risk calculus around AI-assisted academic dishonesty than exists in continuous assessment systems typical of European universities.