Share
On 2 November 2026, the Royal Gazette published the Notification of the Personal Data Protection Committee Re: Guidelines on Personal Data Protection relating to the Development and Use of Artificial Intelligence Technology B.E. 2569 (2026). The Notification was issued under Section 16(3) of the Personal Data Protection Act B.E. 2562 (2019) (“PDPA”), with the Guideline (version 1.0) annexed. Businesses that use AI with customer or employee data should note the following seven points.
1. What goes into and comes out of AI can be personal data
Text typed into an AI tool (prompts), usage records (logs), vector data created by the system (embeddings) and AI-generated results (outputs) can all be personal data if they enable a person to be identified. The PDPA defines personal data as “information relating to a person which enables the identification of such person, whether directly or indirectly” (PDPA, Section 6).
2. A business that deploys AI is generally the data controller
The business decides which data goes into the AI system and how the results are used, so it is the data controller. An AI service provider acting on the business's instructions is a data processor (Section 6). If the provider uses the data beyond those instructions, for example to train its own model, the PDPA deems it a data controller for that use (Section 40, paragraph two). The business must also put in place an agreement with the provider (Section 40, paragraph three).
3. Employees entering data into public AI tools is high-risk
The Guideline treats employees entering customer data into public AI tools as processing on behalf of the organisation, and as high-risk. Businesses should adopt an AI use policy, train staff and use enterprise AI systems backed by data protection agreements (Guideline, Section 2.3, Chapter 7 and FAQ 4.1). The data controller must also provide appropriate security measures (Section 37(1)).
4. Consent is not required every time, but a lawful basis is
The PDPA prohibits collecting personal data without consent unless an exception in Section 24 applies, such as where it is necessary to perform a contract to which the data subject is a party, or necessary for the legitimate interests of the data controller, unless those interests are outweighed by the data subject's fundamental rights in their personal data (Section 24(3) and (5)). Businesses therefore need a lawful basis for each AI use. Data listed in Section 26, such as health or biometric data, is subject to the stricter conditions in that Section.
5. Using an overseas AI provider: check who can access the data
Using an AI service with data centres abroad is not always a cross-border transfer. The Guideline looks at whether anyone else abroad can access the data and whether technical measures or legal terms support that restriction (Guideline, Section 5.3). If the data is in fact sent or transferred abroad, the business must comply with Section 28 or Section 29, as the case may be.
6. AI making important decisions should have a real human reviewer
Where decisions made with AI have legal effects on, or significantly affect, individuals, such as credit approval or staff selection, the Guideline recommends a data protection impact assessment (DPIA) before use, advance notice to data subjects, and a reviewer with real authority to overturn or correct the result, not one who simply rubber-stamps it (Guideline, Sections 5.4 and 6.2).
7. Data breaches and data subject requests: what to do
When a personal data breach occurs, the data controller must notify the Office of the Personal Data Protection Committee without delay and, where feasible, within 72 hours after becoming aware of it, unless the breach has no risk of affecting the rights and freedoms of persons. Where the risk is high, the data subject must also be notified, together with remedial measures, without delay (Section 37(4)). Data subjects may still request access or erasure, or object to processing, as provided by law, even when their data sits in an AI system (Sections 30, 32 and 33).
Note: The Guideline describes itself as preliminary recommendations without legal force, but obligations under the PDPA continue to apply as usual.
What businesses should start doing
• Adopt an AI use policy with clear prohibitions, and train employees.
• Sign an agreement with each AI provider that prohibits using the data beyond what was agreed.
• Carry out a DPIA before using AI for decisions that have legal effects on, or significantly affect, individuals.
• Prepare logging and an incident response plan so breaches can be reported within 72 hours.
References
1. Notification of the Personal Data Protection Committee Re: Guidelines on Personal Data Protection relating to the Development and Use of Artificial Intelligence Technology B.E. 2569 (2026), Royal Gazette, Volume 143, Special Section 261 Ngor, Page 24, dated 2 November 2026 (B.E. 2569), signed on 22 October 2026, with the Guideline (version 1.0) annexed. https://ratchakitcha.soc.go.th/documents/136419.pdf
2. Personal Data Protection Act B.E. 2562 (2019), Sections 6, 16(3), 24, 26, 28, 29, 30, 32, 33, 37 and 40.
English titles and quotations of Thai law in this version are unofficial translations. The Thai text published in the Royal Gazette prevails.
This article provides general information only and does not constitute legal advice for any specific case.