THE KAUTILYA TRUST TEST (KTT) v1.0: A FRAMEWORK FOR CERTIFYING TRUSTWORTHY AI FOR CRITICAL APPLICATIONS BASED ON ARTHASHASTRA 1.10 UPADHA PARIKSHA
Year: 2026
Type: preprint
Abstract: India's AI Mission aims to deploy AI in critical sectors including healthcare, judiciary, finance, andpublic service delivery. The key challenge is certification of trustworthiness. Current globalbenchmarks evaluate AI capability (MMLU, HELM) but not integrity under adversarial temptation.This paper proposes the Kautilya Trust Test (KTT) v1.0, India's first indigenous certificationframework for Trustworthy AI, derived from Kautilya's Upadha Pariksha (Arthashastra, Book 1,Chapter 10). Upadha Pariksha is the world's first documented integrity testing framework comprisingfour tests: Dharma Upadha (adherence to righteous duty), Artha Upadha (resistance to materialinducement), Kama Upadha (resistance to sycophancy and manipulation), and Bhaya Upadha(resilience under coercion).KTT v1.0 translates these four classical tests into 40 measurable scenarios for modern AI systems andproposes a quantitative KTT Trust Score and a three-tier certification model (KTT-Basic, KTTCritical,KTT-Sovereign) suitable for adoption by MeitY, BIS, and NITI Aayog as part of IndiaAIGovernance Guidelines. This is a conceptual framework paper intended to establish priority and invitemulti-stakeholder consultation. SSRN: https://papers.ssrn.com/abstract=7449563
Author: MURSHED, MD FAZLE
Language: English
Subfield: Safety Research
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7212366455
Cites:
Cited by:
Related to: