The Electronic Transactions Development Agency (ETDA) under the Ministry of Digital Economy and Society (MDES) of Thailand released the Draft Act on Artificial Intelligence (Draft AI Act) for public hearing during 9 July to 14 August 2026.  Below is our note on the key provisions of the Draft AI Act.

1.  The Draft AI Act, once passed as a law, will apply to the development, deployment, or any act related to AI systems that affects persons in Thailand no matter the act is carried out in or outside Thailand.

2.  The Draft AI Act does not apply to use of AI by a natural person for personal or household purposes, AI research by state agencies and higher education institutions with the approval of a human research ethics committee, and the research, testing, and development of AI systems before they are placed on the market.

3.  A person may apply to the designated office (Office) under the MDES to test an AI system in a regulatory sandbox in case (1) there are no licensing rules for the AI system, (2) existing rules cannot be applied, or (3) that AI service system is new to the market.  The Specific Committee on Driving the National Artificial Intelligence Plan (Committee) may exempt the participant from certain laws and regulations during the test.  But the participant remains liable for damage caused to third parties.  If the test shows that a law is an obstacle to the development of AI, the law may be amended by a Royal Decree after the draft amendment has been submitted to the House of Representatives and the Senate for at least 60 days.

4.  The Big Data Institute will establish a data sharing center for the development of AI systems.  The Committee may order state agencies to send data to the center.  The Committee may also allow the participant in the sandbox to use personal data collected by state agencies for training and testing AI systems in specific areas such as public health, environment, energy, transport, public services, and target industries, only where anonymized or synthetic data is not sufficient.  The participant must have measures to protect the rights of the data subjects under the Personal Data Protection Act B.E. 2562 (2019).

5.  No person may use an AI system that uses hidden techniques to impair a person’s ability to make his or her own decision, or that unfairly discriminates against people at large by processing data that is not directly relevant to their qualifications for basic rights.  The Committee may announce other prohibited risks.

6.  A Royal Decree may classify types of AI systems as high-risk AI where the use of AI affects national security, health, environment, energy, telecommunications, transport and logistics, or public utilities.  The Office may issue guidelines on risk management.

7.  A foreign provider of high-risk AI that serves users in Thailand must appoint a local coordinator in writing.  However, for specific types of AI systems designated in the Ministerial Regulation, the foreign provider must appoint a local representative authorized to act for the provider without any limitation of liability.  The private sector may propose industry standards on high-risk AI to the Office for approval.

8.  Developers of AI systems that can create or alter images, sound, or video that may mislead the public must assess the risks of misuse, put in place risk mitigation measures, and support machine-readable marks that identify the content as AI-generated.

9.  A person who posts AI-generated images, sound, or video on a publicly accessible computer system must disclose that the content was created or altered by AI if the content may mislead the public on matters such as threats to national security or public safety, statements or acts of election candidates, reliability of regulated investments, properties of food or drugs, or impersonation of another person.  However, a motion picture, video, or program of an artistic, creative, satirical, fictional, or similar nature with AI-generated shall be disclosed in a manner that does not obstruct the viewing and listening of such work.

10.  Platform operators must assess the risks of AI-generated content on their platforms, provide a channel for users to declare and report AI-generated content, verify machine-readable marks, and label AI-generated content clearly before or at the time the user accesses it.  Developers, deployers, and platform operators must keep records of their compliance and review their risk measures at least once a year.

11.  A person who violates this law and causes damage to another person is liable to pay compensation, whether or not the act was intentional or negligent, unless the damage was caused by force majeure or by the injured person, or the act was done under a lawful order of an official.

12.  Administrative fines are up to THB1 million, THB3 million, or THB5 million, depending on the provisions violated.

To see the archive of our past newsletters and articles please click here.