Amid growing global discussions around AI safety and governance, Thailand is strengthening its AI governance efforts through a national “AI Governance Sandbox” to test and assess AI systems before they are deployed in real-world settings.
The initiative is being supported by the country’s Electronic Transactions Development Agency (ETDA), which is expanding the role of its Artificial Intelligence Governance Practice Center (AIGPC) to help organisations identify operational, ethical and public safety risks associated with AI adoption, according to an official announcement.
ETDA identified one of the main challenges in 2026 as not whether organisations were adopting AI, but whether they had sufficient understanding and governance capacity to manage the associated risks. The agency said organisations across Thailand continue to display varying levels of preparedness, particularly in areas such as “accountability, risk assessment and oversight”.
ETDA grouped these challenges into three areas described as “G-A-P”: governance knowledge and understanding, AI readiness, and the ability to translate ethical principles into operational practice.
The agency noted that many organisations still equate AI primarily with chatbots, without fully assessing how AI systems can affect citizens, public services or decision-making processes.
The new sandbox framework is intended to support the practical testing of AI solutions and policy approaches in controlled environments. “The objective is to identify risks before systems are introduced at scale, allowing organisations to refine governance processes, establish evaluation criteria and improve accountability structures,” the announcement said.
“The goal of AI Governance is not to prohibit the use of AI, but to establish risk management systems suited to real-world use, with clear accountability and confidence for citizens and businesses to use AI safely and effectively,” said Rojana Lamlerd, Adviser at ETDA and Director of AIGPC.
Under the proposed model, organisations will be able to submit AI use cases for testing within the sandbox environment. ETDA plans to coordinate participation among system owners, technology developers, specialists and relevant agencies to jointly assess risks and governance requirements.
The initiative also includes plans for cross-border “AI Incident Exercises” to prepare agencies for risks linked to AI-enabled fraud and misinformation. ETDA highlighted concerns over AI-generated scams involving manipulated images, voices and messages that impersonate public authorities or individuals. Such incidents could affect multiple ASEAN countries simultaneously, requiring coordinated information-sharing and response mechanisms.
The initiative aligns with broader regional efforts on responsible AI oversight, including regional cooperation on trusted AI governance. ETDA said collaboration with Singapore is planned to develop Thai-language testing prompts capable of evaluating how AI systems respond to culturally sensitive or high-risk situations.
The planned testing scenarios include situations involving self-harm, child safety and potentially harmful guidance generated by chatbots. ETDA said systems that pass general technical benchmarks may still fail to address the nuances of the Thai language, culture and social context, highlighting the need for locally adapted testing frameworks.
ETDA said that during 2026, it worked with more than 140 organisations across 20 ministries to establish foundational AI governance resources, including guidelines, toolkits and an AI impact assessment playbook. The agency also conducted testing related to ethical large language models in cooperation with the banking sector. ETDA also aims to develop AI governance skills for more than 120,000 personnel.
Also Read: Google DeepMind Backs 16 Green AI Projects; Four Indian Firms Selected
