AISG and Google Host Third Languages Summit for SEA Language AI

The third Languages Summit, co-hosted by AI Singapore, Google, and Thailand's VISTEC, was held in Singapore to address one of the most persistent challenges in natural language processing: the severe under-representation of Southeast Asian languages in large language models (LLMs). The event convened linguists, AI researchers, and technologists from across the region.

Languages such as Thai, Vietnamese, Burmese, Khmer, and many of the Philippines' and Indonesia's regional tongues remain poorly served by mainstream LLMs, which are overwhelmingly trained on English, Mandarin, and a handful of other high-resource languages. This creates a digital divide where millions of SEA users cannot interact with AI tools in their native languages, limiting the technology's reach and practical utility.

The summit focused on concrete strategies: expanding training datasets for low-resource languages, developing evaluation benchmarks that reflect regional linguistic diversity, and fostering cross-border collaboration on tokenisation and model fine-tuning. AI Singapore has been a leading voice on this issue, having previously launched Southeast Asian languages initiatives including the SEA-LION model family.

Why it matters for Singapore: As a multilingual hub at the centre of Southeast Asia, Singapore stands to benefit disproportionately from more inclusive language AI. The republic's own linguistic diversity — spanning English, Mandarin, Malay, Tamil, and numerous Chinese and Indian dialects — makes it a natural testbed for multilingual models. Stronger representation of regional languages also strengthens Singapore's position as a gateway for AI deployment across ASEAN, where hundreds of millions of potential users remain underserved by current technology.