Southeast Asian LLM, AI deployment initiatives among key announcements at ATxSummit 2025

An overview of the 5th Asia Tech X Singapore Summit, covering recent developments, initiatives and AI releases. 

Credit: IMDA
Share this article

Forward-thinking AI solutions were the stars of the show at the 5th Asia Tech X Singapore Summit (ATxSummit), held from 28 to 29 May at Capella Singapore. This year’s theme was Redefining Tech for A Better Future, in line with the global AI revolution’s fast-paced advancements and evolution.

With speakers from over 12 different countries and organisations, three main points were addressed at the summit, being the real-world deployment of AI, notable advancements in AI technology and the testing of AI tools. 

Real-world deployment

Among the most exciting things announced at the ATxSummit included updates on multiple AIs tailored to a diverse range of use-cases, the most notable being the release of the regional-language large language model (LLM) MERaLiON. 

MERaLiON (Multimodal Empathetic Reasoning and Learning in One Network) is an LLM developed through the National Multimodal Large Language Model Programme supported by IMDA, which has previously also supported the release of the text-based LLM Sea-Lion. A $70 million dollar initiative, the Programme was released to increase Singapore’s AI capabilities in the region, fostering talent in a fast-growing regional industry that seeks to join the likes of the United States’ Silicon Valley and China. 

The MERaLiON information booth at ATxSummit 2025.

Credit: IMDA

Previously released in December 2023, Sea-Lion (South-east Asian Language in One Network) is an open-source LLM trained on Southeast Asian languages, designed to take into account the region’s linguistic diversity and range of cultural nuance. Trained on 11 languages in the region, its release was a milestone for language inclusivity in the field of AI, filling in cultural gaps where larger Western models like ChatGPT-4 and Llama have stumbled. 

MERaLiON is, in many ways, a bigger and better version of its predecessor. As a multimodal LLM, it is able to process images, audio and video in addition to text. Its first version, released in December 2024, gained over 90,000 downloads globally from interested parties spanning a diverse range of fields, including academia, startups, and businesses. 

At ATxSummit 2025, MERaLiON’s Version 2 was launched, with key upgrades to its capabilities. These include code-switching capabilities, upgrades to emotional intelligence and paralinguistic understanding, and a wider coverage of Southeast Asian languages. For a model built to cater to people in one of the most culturally and linguistically diverse regions in the world, these upgrades are a key step-up in paving the way for intuitive and culturally aware AI applications in the region. 

Furthermore, A*Star I2R and IMDA have partnered to launch the new MERaLiON Consortium, aimed at bringing together leading experts, end users and industries together to accelerate the learning and adoption of practical AI applications. This move will likely help consolidate talent, ensuring that Singapore remains relevant in the AI revolution even at its fast-developing pace. 

Two new field-specific AIs were also introduced at the summit. Co-developed with SAL, GPT-Legal is a domain-specific LLM aimed at reducing the time taken for lawyers to research case summaries through its integration with LawNet, Singapore’s platform for legal case studies. A trial test allowed it to generate more than 15,000 case summaries, with research time reduced to 10 minutes from 2 days. 

In the medical field, AI scientists and clinicians worked together to develop Enigma, Singapore’s agentic healthcare AI, which is able to streamline administrative processes. It has been piloted at SingHealth institutions to great success, with a 90% reduction in audit time reported and a significant reduction in risk of human error. 

Local and global AI advancements

On 28 May, IMDA announced that “The Singapore Consensus on Global AI Safety Research Priorities” (The Singapore Consensus) will be presented at the Ministerial Roundtable. Built on scientific evidence, the Consensus is meant to bridge the gap between technical information and government policies, allowing policymakers to better understand how to guide the responsible development and use of AI. 

Globally, further streamlining has been done to standardise AI governance across regions. IMDA’s AI Verify Testing Framework has been enhanced in order to better align with US NIST standards, providing a regulatory framework that applies to both markets. 

Testing and refinement of AI tools

In line with the growing global necessity for safe AI prototyping and testing, the Global AI Assurance Pilot was launched in February 2025. Aimed at consolidating best practices and norms regarding the testing and application of generative AI, the initiative was the result of collaboration between the AI Verify Foundation (AIVF) and IMDA. 

Since its inception, the pilot has paired together specialised AI testers and deployers of real-world generative AI applications from 10 different industries. Key findings from the pilot include the observation that many generative AI risks were context-dependent, varying depending on factors like use cases, industries, languages and organisations. The insights gained from the pilot, along with its recommended guidelines and frameworks, will likely influence global norms for AI. 

In addition to the pilot, IMDA has announced the production and release of a Starter Kit, available to companies who wish to test the feasibility, reliability and safety of AI models in business contexts. The Testing Starter Kit for Gen AI Applications is another addition to the arsenal of best practices and methodologies that interested deployers can rely on. 

President Tharman Shanmugaratnam giving his opening speech at the ATxSummit 2025’s opening gala.

Credit: IMDA

“We should aim to develop multilateral governance over AI. That’s a  journey that has begun but takes time.” President Tharman Shanmugaratnam said in his speech at the ATxSG’s opening gala. “In the meantime, we really have to accelerate the building of coalitions of the willing.” Though they are undoubtedly powerful, the steps that Singapore is currently taking to cement itself as a regional AI hub must be strengthened for it to compete with global superpowers on a worldwide scale.

Share this article