The Second-Generation Threat: AI-Assisted Decentralized Militancy and Counter-Extremism Governance in South Asia

Abstract

Existing counter-extremism scholarship and governance practice remain largely anchored to the Islamic State threat model of 2014–2018, which assumed centralized production, stylistically recognizable output, and predominantly Arabic and English distribution. Whether that framework is still adequate for the South Asian threat environment is the question this article addresses. Generative AI tools have, since approximately 2019, enabled decentralized Militant Islamist networks to produce vernacular-language propaganda at near-zero marginal cost, a qualitative shift that the post-2014 counter-extremism consensus did not anticipate. Drawing on a structured comparative case study of three national sub-systems, Pakistan’s Tehreek-e-Taliban Pakistan and its Umar Media wing, Bangladesh’s Ansar al-Islam ecosystem, and India’s WhatsApp distribution environment, and applying Marc Sageman’s network theory of leaderless jihad alongside Tarleton Gillespie’s platform governance framework, with J. M. Berger’s account of algorithmic amplification as a bridging device, the analysis identifies three compounding structural deficits in the current governance architecture: a South Asian vernacular language moderation gap, an absence of definitional categories for AI-generated content in national regulatory frameworks, and inter-state coordination failures shaped by competing geopolitical interests.