LIRNEasia hosted a roundtable discussion on AI, Media and Misinformation on 3 November 2026 in Colombo, bringing together researchers, media professionals and practitioners to examine the role of data and AI in understanding information flows, strengthening the media and information ecosystem, and exploring the opportunities presented by generative AI to counter misinformation narratives spread on social media.
This closed-door event was followed by a panel discussion opened to the wider AI community in Sri Lanka exploring how Large Language Models (LLMs) can be developed to support local languages. The panel discussion brought together experts from Sri Lanka, India, Singapore, and South Africa.
Roundtable Discussions
The event commenced with opening remarks by Helani Galpaya (CEO, LIRNEasia), setting the context for the day’s discussions. Helani provided an overview of the information ecosystem, explained the concept of information disorder, and highlighted LIRNEasia’s research and interventions, including the development of technical tools to address challenges arising from information disorder. This included building and publishing Sinhala and Bengali language corpora (including colloquial language and local news) and an AI-powered support tool for Sri Lankan fact–checkers.
The first session, “Tracking Information Flows – Practical Tools and Responses”, featured presentations by LIRNEasia and partners. During the session, LIRNEasia researchers presented models developed to identify and intervene in the spread of viral misinformation on Twitter in Sri Lanka. Another presentation shared methodological insights into network analysis and digital investigation from Southeast Asia, which helped track the economy of influence work and online manipulation. The research suggested that short, less source-focused misinformation produced by smaller accounts tended to go viral. The models may have practical applications in warning users about already debunked claims and identifying early warning points.
The second session was titled “Strengthening the Media and Information Ecosystem”. LIRNEasia researchers shared findings from the study of narrative convergence in the context of increasingly concentrated media ownership in Sri Lanka by focusing on reporting on Cyclone Ditwah in 2025. This study utilized natural language processing techniques to conduct sentiment analysis, stance detection, and framing analysis. The research suggested that state-owned newspapers were more favorable to the government, but other variations were difficult to explain. There were also noticeable differences between Sinhala and English language reporting, even within the same ownership structure. One of the key limitations of the research was that Tamil language newspapers were not analyzed. Discussions also covered a partnership to improve ethical news reporting in Sri Lanka using an AI enabled system. This involved the use of Large language Models to rank daily scraped articles in order of risk containing unethical reporting, helping reviewers prioritize their time.
During the third session, “Risks, Realities and Opportunities with Generative AI”, researchers shared the results of an experiment comparing the effectiveness of agentic fact–checking against a more controlled pipeline of LLM assisted fact–checking, and a pilot study on the perception of fact checks labelled as AI-generated. In the subsequent discussion, participants exchanged insights on information flows, media resilience, misinformation, and the role of generative AI in addressing information disorders as well as contributing risks.
Panel Discussion

Helani Galpaya (left) presents opening remarks at the evening event ‘A Panel Discussion on Building Local Language AI Capabilities’, which was followed by the keynote address by Romesh Ranawana, PhD (right).
The evening public event commenced with the keynote address by Romesh Ranawana, PhD (Chair, AI Advisory Committee, GovTech), who provided an overview of Sri Lanka’s current standing and the AI models and datasets developed to date. Romesh also emphasized the importance of high-quality datasets and robust AI models. The keynote was complemented by two presentations by Shashank Bharadwaj (Lead – Revenue Strategy & Operations, Founder’s Office, Sarvam AI) and William Tjhi, PhD (Deputy Director – AI Products at AI Singapore), who showcased regional approaches to developing AI models from a private sector organization in India and a Southeast Asian language project.

Shashank Bharadwaj (left) and William Tjhi, PhD (right) present insights into regional approaches in developing AI models for local languages.
The presentations were followed by a panel discussion which identified several critical points which Sri Lanka needs to address to meet its AI ambitions.

From left to right: Merl Chandanda, Shashank Bharadwaj, William Tjhi, PhD, Seani Rananga, PhD and Ruvan Weerasinghe, PhD engaging in the panel discussion.
The discussion was moderated by Merl Chandanda (Senior Research Manager and Data Algorithms Policy Team Lead, LIRNEasia) and brought together Ruvan Weerasinghe, PhD (Dean of Research, Informatics Institute of Technology (IIT), Sri Lanka); William Tjhi, PhD (Deputy Director – AI Products at AI Singapore); Seani Rananga, PhD (Lecturer and Researcher, Masakhane Project and the African Institute for Data Science and Artificial Intelligence (AfriDSAI)), University of Pretoria, South Africa); and Shashank Bharadwaj (Lead – Revenue Strategy & Operations, Founder’s Office, Sarvam AI).
A common thread that emerged during the discussion was the need for collaboration among multiple stakeholders such as universities, researchers, government, and industry. Researchers need to develop more robust benchmarks and make datasets more open and accessible; application developers should work closely with researchers to translate findings into practical solutions; the government has a role in establishing standards and providing funding for such initiatives, and communities should have a voice in shaping the development and deployment of these technologies.

