Thailand's Draft AI Law: What Companies Should Do Now

Where Thailand's draft AI law stands as of November 2026, what it proposes, which rules already apply, and seven practical steps for companies, with sources.

Thailand does not yet have an AI law in force. A draft AI Act was released for public consultation in July 2026, and the consultation has closed. Until a final law is passed, Thai companies are governed by existing laws such as the Personal Data Protection Act (PDPA) and by sector guidelines. The sensible move now is to prepare: list the AI you use, note where personal data flows, and keep records of how systems are used and checked.

This article is for legal, compliance, IT and operations leads at Thai companies. It sets out what we could verify from published sources as of 13 November 2026, what the draft proposes, and a practical list of steps. It is general information, not legal advice. Confirm your position with qualified counsel, and check the Electronic Transactions Development Agency (ETDA) for the latest status.

What is Thailand's draft AI law?

Thailand's draft AI law is a proposed Act, prepared under the ETDA and the Ministry of Digital Economy and Society, that would set rules for developing and using AI in Thailand. It follows draft principles put out for public hearing in June 2025. A fuller draft Act was released for consultation in July 2026, with the hearing closing in August 2026. As of the sources we read, it has not been enacted.

It proposes a risk-based approach: some AI would be prohibited, some designated high-risk and subject to stricter duties, and some subject to transparency rules, for example for chatbots and AI-generated content. It would apply to developers, providers and deployers, including those outside Thailand whose AI affects people in Thailand. Details can change before any final text is passed, so treat every point as a proposal and not a rule.

Where does the law stand today?

Here is the timeline we could confirm, with sources.

  • June 2025: The ETDA and the ministry held a public hearing on draft principles of an AI law, according to Norton Rose Fulbright and Tilleke & Gibbins.
  • March 2026: Two law firm reviews, by Baker McKenzie and by Tilleke & Gibbins (linked above), still described the framework as a draft under development, with no unified AI statute enacted.
  • July 2026: The ETDA released a new draft Act for public consultation. Tilleke & Gibbins gives the date as 2 July, while Baker McKenzie gives 9 July and says the hearing closed on 14 August 2026. The sources disagree on the release date, so we say only "July 2026".

We found no official confirmation that the Act has been passed since the hearing closed. Enactment timing is not stated in the sources we opened, so treat any date you see quoted as an estimate.

What would the draft Act require?

The points below come from the law firm summaries linked above. They describe a draft, so the final text may differ.

  • Risk categories. Baker McKenzie describes prohibited, high-risk, licensing-required and transparency-obligation categories. The transparency category covers deepfakes, chatbots and generative AI.
  • Duties for high-risk AI. According to Tilleke & Gibbins, high-risk systems would be designated by royal decree and would include areas such as health, energy, telecommunications, transport and public utilities. Deployers would keep operating logs for a minimum period, which it says is six months in the Thai text.
  • Labels for AI content. The same article says developers of image, audio and video generators must embed machine-readable marks, and platforms must label AI content.
  • Reach beyond Thailand. The draft applies to conduct affecting people in Thailand, even if it happens abroad.
  • Liability and enforcement. Baker McKenzie notes strict and joint liability, and powers to suspend services, order recalls and block AI systems.
  • Government vendors. Baker McKenzie also reports that vendors serving government agencies and critical information infrastructure could face mandatory contract terms and data localization.
  • Sandbox. Tilleke & Gibbins reports a regulatory sandbox framework.

What rules already apply to AI use?

Without an AI statute, companies still face obligations. Baker McKenzie's March 2026 review lists exposure under intellectual property, personal data protection, consumer protection, cybersecurity and computer crime laws. It also gives examples of sector rules: disclosure of AI use in court filings, labels on AI-generated advertising, and the Bank of Thailand's AI risk-management expectations.

Tilleke & Gibbins lists AI guidelines from the Bank of Thailand, the Securities and Exchange Commission, the Office of Insurance Commission and the National Cyber Security Agency. It notes that some guidelines carry no legal binding effect, but regulators expect compliance. It also reports that a draft Personal Data Protection Committee guideline on personal data in AI development and use was released in February 2026.

If your AI touches personal data, the PDPA is the law that matters most today. Our guide to PDPA and AI data residency explains where data can live.

What should companies do now?

You do not need a final law to start. Most of the work below is good practice under any version of the rules.

  1. List your AI systems. Include tools staff use on their own, and AI inside software you buy.
  2. Map the data. Note which systems handle personal data and where it is stored and processed.
  3. Rank by impact. Flag anything that affects health, money, access to services or safety. These resemble the areas the draft names as high-risk.
  4. Keep records. Log how each system is used, who approves changes, and how you test it. Logs are one of the duties the draft describes.
  5. Name an owner. Someone should decide what is allowed, and who handles a serious error.
  6. Check your vendors. Ask where they host data, who can access it, and how they would meet new duties.
  7. Plan for human review. Keep a person responsible for decisions that affect people.

Baker McKenzie's March 2026 review reaches a similar conclusion: build a formal AI governance model and internal AI usage policies, and update documents and records.

How does the draft connect to sovereign AI and public procurement?

If you sell to government, two parts of the draft matter. Baker McKenzie's note on data localization and mandatory contract terms for government vendors affects where you may run systems and how you contract. We explain the concept in what sovereign AI is and whether your company needs it.

For agencies, our list of AI compliance questions in public procurement is a starting point for tenders. Because the draft may change, ask vendors how they would adapt, and avoid contract terms that lock in a design built around one reading of a draft.

What is not yet known?

Be honest about the gaps. We could not verify the final text, the date of enactment, or which systems will be designated high-risk, because designations would come through later legislation. Sources also differ on the July 2026 release date.

Do not rebuild a working system around a draft. Do not wait for the law before you tidy your records either. The two cheapest things you can do are to know what AI you run and to know where its data goes.

How to start

  1. Run a one-week inventory of AI systems and AI features in your software.
  2. Mark the ones that touch personal data or affect people's access to services.
  3. Write a one-page AI use policy and name an owner.
  4. Ask your top vendors where data is hosted and how they log use.
  5. Set a reminder to check the ETDA and your sector regulator each quarter.

UPSPIRE's solution demos are concepts built with synthetic data, and our team is not a law firm. If you want help mapping your AI and data before the rules settle, contact our team.

Frequently asked questions

Does Thailand have an AI law in force?

Not as of the sources we read. A draft AI Act was released for consultation in July 2026 and the hearing has closed. Existing laws such as the PDPA and sector guidelines still apply. Check the ETDA for the latest status.

Who wrote Thailand's draft AI law?

It is prepared under the Electronic Transactions Development Agency (ETDA) and the Ministry of Digital Economy and Society, following draft principles released for public hearing in June 2025.

Would the draft apply to companies outside Thailand?

According to law firm summaries, the draft covers conduct affecting people in Thailand even if it happens abroad. Because it is a draft, the final text may differ, so confirm with counsel.

Do we need to wait for the final law before acting?

No. Listing your AI systems, mapping personal data, keeping usage records and naming an owner are useful under existing laws and likely under any final version.

Is this legal advice?

No. It is general information from published sources. Confirm your own position with qualified counsel and the relevant regulator.

UPSPIRE Team

AI consulting and infrastructure, Bangkok. UPSPIRE's consultants and engineers build AI agents, Thai-language knowledge systems and sovereign AI infrastructure for enterprises in Thailand and Southeast Asia.

Ready to put this to work?

Turn this playbook into a working system

From a first agent to a national platform, the same team plans it, builds it and runs it with you.

Book a consultation