Overview

Thailand has taken a significant step towards regulating artificial intelligence with the publication of the Draft Artificial Intelligence Act (the Draft AI Act) by the Electronic Transactions Development Agency (the ETDA), under the Ministry of Digital Economy and Society. The Draft AI Act was published for public consultation on 9 July 2026, with comments due by 14 August 2026. Following the consultation period, the ETDA will review the feedback received before proceeding with further legislative steps.

The Draft AI Act consolidates two earlier proposals: (i) a 2022 draft which adopted a risk-based approach modelled on the EU Artificial Intelligence Act; and (ii) a 2023 ETDA draft focused on promoting AI innovation through sandboxes, data sharing, standards and risk assessment. If enacted, the Draft AI Act would establish Thailand's first comprehensive legal framework for AI, supporting innovation while promoting trust, transparency and the responsible use of AI.

Scope

Territorial scope

The Draft AI Act applies to the development, deployment and other activities involving AI systems that affect persons in Thailand, regardless of where such activities take place. This broad and potentially extra-territorial jurisdictional reach of the Draft AI Act may pose extra-territorial compliance obligations for foreign AI providers and deployers whose systems affect users or persons in Thailand.

Persons and entities regulated

The Draft AI Act primarily regulates two categories of actors:

  1. Providers: persons who sell or provide AI systems under their own name or trademark, whether or not for consideration; and
  2. Deployers: persons using AI systems, subject to limited exceptions for personal or household use.

In addition to Providers and Deployers, the Draft AI Act also imposes specific obligations on other participants in the AI value chain - these include users, content publishers, platform operators and government agencies, depending on the relevant AI system and the nature of their activities.

Exemptions

The Draft AI Act excludes certain activities from its scope, including: (i) personal or household use of AI systems; (ii) AI systems developed for academic research by authorised government research agencies or private higher education institutions, subject to ethical research committee approval; and (iii) AI research, testing or development activities conducted before commercialisation. Further exemptions may be prescribed by future Royal Decrees.

Key implications

Innovation-focused framework

The Draft AI Act seeks to promote AI development through mechanisms such as AI innovation sandboxes, government data sharing, certification schemes, AI support services and industry self-regulation.

Risk-based regulatory framework

The Draft AI Act adopts a risk-based approach by classifying AI systems into three categories:

  1. AI systems posing unacceptable risks

    Certain AI systems are prohibited outright, including systems that: (i) use subliminal or manipulative techniques to distort decision-making in a manner that causes significant harm; or (ii) cause unfair discrimination in accessing fundamental rights through the processing of irrelevant or unjustified data.

    Separately, where it is considered necessary to maintain public safety, financial and commercial stability, or prevent public harm, a Royal Decree may be passed to impose notification, registration or licencing requirements o the development or deployment of certain AI systems.
  2. High-risk AI systems

    AI systems used in areas such as national security, healthcare, environment, energy, telecommunications, transport, logistics, and public utilities, are subject to enhanced governance requirements.

    Providers of high-risk AI systems are required to implement governance measures proportionate to the level of risk posed by their systems. These include requirements relating to risk management, transparency, technical documentation, and the protection of personal data.

    Deployers are likewise required to implement appropriate governance measures, including maintaining system logs, monitoring AI performance and reporting serious risks identified during deployment.
  3. AI systems that generate or manipulate content

    AI systems capable of generating or altering images, audio, video or other content in a manner that may mislead the public—including generative AI, chatbots and deepfakes—must comply with specific transparency obligations to ensure that users are informed when interacting with AI-generated or AI-altered content. Developers of such systems must assess the risks of deception, identity fraud and content manipulation, implement proportionate risk-mitigation measures, and embed machine-readable marks to enable the identification of AI-generated content.

    Platform providers hosting user-generated content must, where applicable, assess content-related risks, verify provenance information using machine-readable marks, clearly label AI-generated or AI-altered content, provide accessible reporting channels for users, and respond promptly to reports of non-compliance. Deployers that use AI to create or modify content for public dissemination must also disclose the use of AI and refrain from using AI for deception, fraud or identity misrepresentation.

Controlled-contract AI businesses

The Specialized Committee under the Draft AI Act may designate certain AI businesses providing services to government agencies or critical information infrastructure as “controlled-contract AI businesses”. This designation enables the imposition of mandatory contractual requirements on matters such as data processing, cross-border data transfers, cybersecurity and termination conditions.

Enforcement, penalties, and strict liability

The Draft AI Act adopts a graduated enforcement framework, allowing regulators to issue corrective orders and, where necessary, seek court orders to suspend AI services, recall AI products or block access to non-compliant AI systems in Thailand.

Non-compliance with the Draft AI Act that results in damage may give rise to liability for resulting losses. Administrative penalties may also apply for specified violations, with fines ranging from THB 1 million to THB 5 million, depending on the nature and severity of the breach.

The Draft AI Act further introduces a strict liability regime for persons who violate its provisions. Such persons may be required to pay joint compensation for damages arising from violations, regardless of whether the conduct was intentional or negligent, subject to certain statutory exceptions.

Next steps and recommendations

The Draft AI Act represents a significant milestone in Thailand’s evolving technology regulatory landscape. As the Draft AI Act progresses through the public consultation and legislative process, the draft provisions may be subject to further revisions based on stakeholder feedback. Businesses developing or deploying AI systems that may affect persons in Thailand may wish to consider the following steps:

  • review existing AI governance frameworks and identify gaps against the Draft AI Act’s requirements, particularly for high-risk AI systems;
  • evaluate transparency and disclosure obligations for AI systems that generate or manipulate content; and
  • monitor the legislative process for further developments, as provisions may change following the public consultation.

We will continue to monitor developments and are available to discuss the potential implications for your business.

We would like to thank Supisara Techakasembundit and Bantita Kraisripotikul for their contribution to this post.

Contact