Imagine an AI system scanning thousands of social media posts in seconds, detecting recurring words, images and narratives, mapping how they spread, and flagging content that may be connected to violent extremism. The technology is impressive. But what happens when the pattern is right, while the meaning is wrong?
A meme might just be a joke, while a phrase that looks harmless on its own could mean something very different in another context. Religious references can be part of everyday conversation, but may also be taken out of context and used to justify violence. For AI, distinguishing between these meanings is difficult because the same words or images can carry different meanings depending on the people, communities and situations involved.
At the heart of SEAN-CSO’s first webinar, “Fellowship Kick Off: Artificial Intelligence and P/CVE,” held on 8 September 2026, was a closer look at the fundamentals of violent extremism and the different factors that shape radicalisation. The discussion then turned to where artificial intelligence could meaningfully support P/CVE efforts. Around 55 participants from across Southeast Asia joined the session, led by Raneeta Mutiara, a P/CVE and public safety specialist whose work focuses on P/CVE, gender, rehabilitation and reintegration, AI, and online radicalisation.
Understanding Violent Extremism in Context
The discussion began with a simple question, “What comes to mind when you hear violent extremism?”
Through Slido, participants shared words including terrorism, radicalisation, ideology, intolerance, grievances, hatred, recruitment, misogyny and harm. The responses reflected the complexity of a phenomenon that does not have one universally applicable definition.
One distinction is particularly important, having a radical or extreme view does not automatically mean someone will engage in violent extremism. Violence remains central to the concept, while radicalisation is better understood as a process shaped by multiple factors, including social relationships, ideology, digital environments, political conditions and individual experiences.
Religion also requires careful consideration. Violent extremism can emerge from religious as well as non-religious ideological environments, but religion itself should not be equated with violent extremism. Extremist actors may misuse, reinterpret or distort religious concepts to legitimise violence.
Understanding these distinctions matters because AI can only be as useful as the problem it is designed to understand.
How AI Is Changing the Threat Landscape
Generative AI has changed the speed, scale and accessibility of content production. Propaganda that once required specialised knowledge, equipment, multiple people and significant time can now potentially be produced by a small number of actors using inexpensive tools.
AI can generate text, images, audio and video, while making harmful narratives easier to personalise, localise and amplify. Such content does not always appear explicitly extremist. It may take the form of memes, entertainment, provocative commentary or seemingly casual interactions, gradually contributing to the normalisation of harmful ideas.
At the same time, AI can support P/CVE practitioners by helping detect patterns, analyse narratives, understand how ideas circulate, develop counter-narratives and evaluate interventions. The challenge is ensuring that these capabilities do not come at the expense of context, accuracy, privacy, human rights and accountability.
Why Local Context Matters
Southeast Asia cannot be treated as a single P/CVE environment. The region encompasses hundreds of languages and dialects, different political histories, digital cultures, local grievances and institutional capacities.
An AI system developed for one country cannot simply be transferred to another and expected to understand its threat environment. Localisation needs to go beyond translation. It must account for language, culture, politics, digital behaviour, ideological movements and local experiences. AI may be able to identify that something is happening. Understanding why it is happening still requires people who know the context.
Building P/CVE Capacity in Indonesia
The Indonesian case study highlighted another challenge. Access to AI training and resources remains uneven. For Raneeta, strengthening P/CVE awareness and AI literacy needs to happen alongside greater government investment and collaboration across sectors.
“The main homework is to first socialise and familiarise P/CVE, and the government has to invest heavily in AI, and only then combine AI and P/CVE, implementing and empowering them so that people have digital and AI literacy. […] When in reality, this is a social issue So the government needs to create or reframe this first so that access is open. People will become aware regardless of religion, social status, etc.”
Raneeta Mutiara
This means P/CVE cannot remain the responsibility of governments or security institutions alone. Communities, CSOs, universities, researchers and practitioners each bring different forms of knowledge, and greater collaboration is essential to developing approaches that are both technically capable and socially appropriate.
Towards Responsible AI for P/CVE
AI can create new risks while also opening new possibilities for P/CVE. It can identify patterns while missing context, increase efficiency while introducing bias, and support decision-making while raising difficult questions about accountability.
For this reason, human judgement needs to remain central. Cultural understanding, ethical considerations and knowledge of the communities involved cannot simply be automated.
The future of AI-enabled P/CVE will depend not only on better technology, but on whether the people using it have the literacy, collaboration and safeguards needed to use it responsibly. Violent extremism does not come with a single face, language, pathway or explanation. Neither should AI.
The conversation continues with SEAN-CSO’s next webinar, “Developing Counter-Narratives with AI,” on 15 September 2026, from 1:30–3:30 PM WIB (GMT+7). The session will explore how AI can support counter-narrative development while keeping context, creativity and human judgement at the centre.
Register for the next webinar:
https://seancso.org/registerseancso2026
SEAN-CSO provides a platform for civil society organisations in Southeast Asia to network and share resources. Through this network, members can support each other, share knowledge, and coordinate efforts to address the region’s social challenges. In doing so, SEAN-CSO contributes to strengthening the voice of civil society at both regional and international levels.