Generative AI and Copyright in India: An Empirical Study of Stakeholder Perceptions on Consent, Licensing, Transparency, Compensation and Legal Reform

Abstract

Generative AI has brought a practical copyright question to the centre of Indian policy: on what terms, if any, should copyright-protected works be used in training models? This paper reports an exploratory mixed-method study of 105 respondents drawn from five stakeholder groups - software professionals, lawyers and academicians, authors, content creators, and students or researchers. Nineteen Likert-scale items and three open-ended questions were used to examine views on permission, licensing, compensation, disclosure, research use, trust, enforcement and legal reform. Mandatory licensing attracted the strongest support (M=4.39; 85.7% agreement), followed by AI-specific legislative reform (M=4.36; 84.8%), legal prohibition of unauthorised training (M=4.34; 82.9%), training-data disclosure (M=4.30; 81.0%) and creator compensation (M=4.29; 78.1%). Seventeen of the 19 items differed significantly from the neutral midpoint after false-discovery-rate correction. The two status-quo items - adequacy of the present Copyright Act and sufficiency of existing enforcement - did not show a clear departure from neutrality. The responses also show that support for creator protection is compatible with more lenient treatment of genuine non-commercial research and with recognition of AI's social benefits. Professional-category differences did not remain significant after multiple-testing correction. The findings should not be read as nationally representative, but they provide stakeholder evidence relevant to the current Indian debate on AI training, copyright licensing, transparency and reform.