Background
AI regulation in Thailand is still in early stages. Currently there are several non-binding AI guidelines for relevant parties (such as developers, researchers, providers, or operators) issued by certain Thai government agencies including National Science and Technology Development Agency (“NSTDA”) and Electronic Transactions Development Agency (“ETDA”). Such AI guidelines aim to encourage relevant parties to assess the risk from using AI in their activities and provide recommendations with regards to fundamental principles; i.e., privacy, security and safety, reliability, fairness and non-discrimination, transparency and accountability, and human oversight. However, these guidelines serve merely as a voluntary recommendation for compliance.
Having said the above, currently, two draft legislations and one draft principle for AI legislation have been introduced as follows:
• Draft Royal Decree on Business Operations that Use Artificial Intelligence System;
• Draft Act on the Promotion and Support of AI Innovations in Thailand; and
• Draft principles for AI legislation.
The details are set out below.
Draft Royal Decree on Business Operations that Use Artificial Intelligence System
Draft Royal Decree on Business Operations that Use Artificial Intelligence System (the “Draft Decree”) was introduced by the Office of the National Digital Economy and Society Commission to be enacted by virtue of the Electronic Transaction Act B.E. 2544 (2001). The Draft Decree was created on a risk-based approach influenced by the EU AI Act pursuant to which the regulatory requirements and supervision for AI systems will be stricter as the risk of such AI system increases.
Pursuant to the Draft Decree, the AI systems are separated into three categories as follows:
| AI Category | Proposed Regulation |
| 1. Unacceptable Risk
(e.g., social scoring system, AI system that can access sensitive personal information such as age or disabilities) |
Prohibits all use with limited exception on a case-by-case basis. |
| 2. High Risk
(e.g., credit scoring, predictive policing system, recruitment software, autonomous vehicle) |
Requires prior registration with the regulator and compliance with risk control and management measures. |
| 3. Limited Risk
(e.g., chatbots, deepfakes) |
Imposes transparency obligation. |
The Draft Decree only regulates AI systems within the reach of the general consumer and does not extend to those undergoing research and development.
The Draft Decree also contains extraterritorial applicability to service providers located outside of Thailand, requiring service registration with the regulator and the appointment of a local point of contact. In terms of liabilities, several measures and liabilities have been proposed, for example, if an AI service provider fails to comply with the relevant obligations, such providers shall be subject to an administrative order, including fine penalty. In addition, criminal liability may be imposed in case which high-risk AI is not registered with the regulator before the provision of service.
Draft Act on the Promotion and Support of AI Innovations in Thailand
Draft Act on the Promotion and Support of AI Innovations in Thailand (the “Draft Act”) was introduced by the Electronic Transactions Development Committee of the ETDA. The Draft Act aims to promote and provide necessary support for the development of an AI ecosystem through legal mechanisms.
The AI promotion committee, consisting of representatives from various sector-specific regulators, is proposed to be established to supervise the implementation of the Draft Act, which will correspond with different practices across different industries and utilizes the following key mechanisms.
| Key Proposed Mechanisms | Key Features |
| 1. Testing Center Sandbox | Allows an AI service provider to:
• test an AI system in a controlled environment prior to launch; • seek the regulator’s recommendations on technical and regulatory concerns; • potentially be exempt from regulations that would otherwise hinder AI development. |
| 2. Standard Contractual Terms for AI service providers | Provide minimum provisions that each AI service provider must provide in its terms and conditions, including:
• obligations and responsibilities of a service provider; • warranty of service standards; • provisions relating to rights in data arisen from or related to AI. |
In terms of liabilities, there is currently no specific sanction outlined in the Draft Act, save for when an AI service provider does not act according to its advertisement relating to the quality of service. In such case, the ETDA will be empowered to issue a corrective order.
Draft principles for AI legislation
Following the release of draft legislations on AI as provided above, the ETDA held a session to explain the draft principles for AI legislation and regulatory approach for an updated AI law (the “Draft Principles”). While earlier versions of the draft legislations were modeled upon the EU framework, ETDA noted the need to revise them to better reflect Thailand’s evolving legal and technological context.
The Draft Principles adopt a risk-based approach. However, instead of listing prohibited or high-risk AI systems directly in the primary legislation, authority would be delegated to an enforcement agency and sectoral regulators to define such categories through subordinate regulations. Providers of high-risk AI would be subject to additional obligations, including implementing risk management frameworks aligned with international standards, appointing local representatives (for offshore providers), and reporting serious incidents. Deployers of high-risk AI would be required to ensure human oversight, maintain operational records, safeguard data quality, notify affected individuals, and cooperate in investigations, with potential liability if they fall below professional standards of care.
The Draft Principles include supportive measures such as permitting certain uses of online data for text and data mining (subject to rightsholder reservations for commercial uses) and establishing regulatory sandboxes in order to promote innovation. Similar to the feature proposed by the Draft Act, such sandboxes would allow real-world testing under controlled conditions, including limited use of personal data for public-interest AI development. Participants acting in good faith can be protected from regulatory penalties during experimentation, though civil liability for damages would still apply.
The Draft Principles also outline general legal principles, including nondiscrimination against AI-generated contracts or administrative decisions, affirmation that actions generated by AI must be attributable to humans, protections against unforeseeable AI errors, and granting individuals the right to explanation and appeal (i.e., the right to understand how AI systems are developed and the ability to appeal decisions made by or with AI).
In terms of an enforcement pursuant to the Draft Principles, enforcement powers include issuing administrative orders to suspend or prohibit high-risk AI services, removing online content, seizing physical products, and coordinating with relevant government agencies to restrict access within Thailand.
Summary
In light of the above, the legislative process for draft AI laws in Thailand remains at an early phase, and the timeframe for their enactment is still uncertain. However, businesses are encouraged to monitor developments and prepare their compliance frameworks accordingly.
Timblick & Partners Ltd.
By: Sukavit Duangploy
Date: 18 February 2026