Phnom Penh – In a significant move to ensure Cambodia’s cultural identity is represented in the next generation of artificial intelligence, Mr. En Sovann, Head of the Department of Digital Government Transformation and chair of the AI and Data science working group at MPTC, attended the 3rd Global Partnership on Artificial Intelligence (GPAI) Innovation Workshop in Tokyo this past May. His participation was pivotal to raise the importance of the development of a multicultural benchmarking dataset for Large Language Models (LLMs), a collaborative project between Cambodia, ASEAN fellows, and Japan’s National Institute of Information and Communications Technology (NICT).

The GPAI workshop, which gathered 130 experts from 35 countries, served as a strategic global venue to advance this vital work.

The Critical Need for a Multicultural AI

Today’s leading AI systems, such as ChatGPT, are built on LLMs trained on trillions of words. However, the vast majority of this data is in English, creating a severe imbalance. This means that without deliberate intervention, these powerful tools may fail to understand the nuances of Khmer language, history, social values, and cultural context. The risk is an AI future where Cambodian identity is diluted or misrepresented, and technology fails to serve its citizens effectively.

Recognizing this challenge, Cambodia’s Ministry of Post and Telecommunications (MPTC) has been proactively collaborating with NICT since January 2025 on a project to build a high-quality dataset that truly reflects our unique cultural and linguistic heritage. The goal is to create a benchmark that can be used to build and evaluate AIs, ensuring they are not only fluent in Khmer but also conscious of its cultural underpinnings.

Advancing the Mission at the GPAI Summit

The GPAI Innovation Workshop provided the ideal platform to elevate this project from a regional collaboration to a global priority. The workshop’s agenda was structured around key themes essential for an equitable AI future, including Open Source AI, AI for the Global South, and Governance Interoperability.

Crucially, one of the four main themes was “Multilingual and Multicultural Aspects of Generative AI.” During these sessions, Mr. Sovann shared insights from the ongoing work with NICT, advocating for a global framework that supports the development of culturally-aware AI.

The Path Forward: Securing Cambodia’s Digital Heritage

By continuing to lead in this area, this project does not end with just building a dataset. We are building a future where technology respects our past, serves our present, and preserves our unique cultural identity for generations to come.