How does the Bank of Thailand regulate AI in banking?

Last updated Nov 20, 2026 · Updated as rules change

The Bank of Thailand (BOT) sets its expectations for AI in banking through a policy guideline rather than a rule: on 12 September 2025 it issued its policy guideline on risk management of the use of artificial intelligence systems, by circular ธปท.ว. 5994/2568, to commercial banks, specialized financial institutions and non-bank payment service providers. The guideline, published in Thai, makes a provider responsible for its AI systems and for decisions based on their outputs, asks the board and senior management to govern AI under responsible-AI principles such as FEAT (fairness, ethics, accountability, transparency), and sets controls for data, models and AI-specific cyber threats such as prompt injection and data poisoning. In consultation the BOT declined to create high-risk or prohibited tiers of AI, leaving risk grading to each provider. Separately, minimum practices on digital fraud issued on 14 September 2026 require the largest banks and significant e-money providers to have technology-based transaction monitoring that ranks risk, with AI and machine learning named as examples, in use within 2026.

Checked 2026-11-20 · Bank of Thailand (BOT), Thailand's central bank and the supervisor of commercial banks, specialized financial institutions and payment service providers
Full nameBank of Thailand (BOT), Thailand's central bank and the supervisor of commercial banks, specialized financial institutions and payment service providers
RoleCentral bank and financial-institution supervisor whose AI position is a principles-based policy guideline issued by circular, with AI named as a monitoring technology in its mandatory digital-fraud minimum practices
Force on banksSupervisory guidance
Applies toCommercial banks registered in Thailand, specialized financial institutions and non-bank electronic payment service providers, the addressees of the BOT's AI risk management guideline of 12 September 2025. The guideline is a policy guideline (แนวนโยบาย) framed as what providers should do and applied by use case; it sets no effective date and no penalties of its own. The digital-fraud minimum practices of 14 September 2026, which name AI and machine learning as monitoring technologies, are mandatory for commercial banks and licensed e-money and electronic money-transfer providers
Key documentPolicy guideline on risk management of the use of artificial intelligence systems (BOT circular ธปท.ว. 5994/2568, 12 September 2025; published in Thai)
Latest move14 Sep 2026: the BOT issued minimum practices for digital fraud management (ธปท.ว. 5713/2569) requiring domestic systemically important banks and significant e-money providers to develop and start using technology-based abnormal-transaction detection within 2026, with AI and machine learning named as examples; 12 Sep 2025: the BOT issued its AI risk management policy guideline (ธปท.ว. 5994/2568) after consulting from 12 to 30 June 2025.
Documents tracked3 · all documents →

The AI guideline came out of a short consultation. A draft, which the BOT titled in English Guiding Principles for Artificial Intelligence Risk Management, was open for comment on the BOT website from 12 to 30 June 2025, and the final policy guideline followed by circular ธปท.ว. 5994/2568 on 12 September 2025, prepared by the BOT's IT risk supervision and examination department. It defines an AI system as one that imitates human intelligence and learns from data, excluding rule-based automation whose steps people define, and it names three families of risk: data, model development and cyber threats, including prompt injection, model inversion, data poisoning and adversarial attacks. It adds to the BOT's existing IT risk rules, its IT risk and third-party risk management implementation guidelines, its data governance policy and its market conduct rules, which it names, rather than replacing them.

The guideline has two parts. Governance makes the board and senior management accountable for AI decisions and actions and asks for an AI policy reviewed regularly, roles across the three lines of defence, staff training against over-reliance on AI and a defined risk appetite; human-in-the-loop or human-over-the-loop oversight is something providers may consider for core functions such as approving loans, account openings and transfers, and for work that directly affects customers. Development and security sets controls for training-data quality and leakage, model evaluation and testing, hallucination in generative AI, explainability, content filtering and attack testing, naming OWASP and MITRE ATLAS as references. In consultation the BOT kept this principles-based design: it did not add high-risk or prohibited categories, and it narrowed a proposed option for customers to disable or bypass AI services into informing customers and possibly offering a route to staff.

AI also appears in the BOT's fraud rules. The minimum practices for digital fraud management issued on 14 September 2026 (ธปท.ว. 5713/2569) under the Digital Fraud Management rules require commercial banks and e-money providers that offer transfers to other providers to upgrade transaction monitoring in line with their fraud risk, naming artificial intelligence and machine learning as example technologies, and require domestic systemically important banks and significant e-money providers to have a technology-based system that ranks the risk of each transaction, account or customer in use within 2026. Providers must also track the accuracy of their mule-account detection and avoid undue impact on legitimate customers. All three BOT documents tracked here are published in Thai.

What has the Bank of Thailand actually published on AI?

DateDocumentStatus
Sep 14, 2026BOT Minimum Practice for Digital Fraud Management — Minimum Practice for Digital Fraud Management (แนวปฏิบัติขั้นต่ำสำหรับการบริหารจัดการภัยทุจริตดิจิทัล)In force
Sep 12, 2025BOT AI Risk Management Policy Guideline — Policy Guideline on Risk Management of the Use of Artificial Intelligence Systems (แนวนโยบายธนาคารแห่งประเทศไทย เรื่อง การบริหารจัดการความเสี่ยงของการใช้งานระบบปัญญาประดิษฐ์)Final
Jun 12, 2025BOT Draft AI Risk Management Guideline — Draft Policy Guideline on Risk Management of the Use of Artificial Intelligence Systems (Guiding Principles for Artificial Intelligence Risk Management)Superseded
DateTypeDocument / event
Dec 2026MilestoneDeadline for D-SIBs and significant e-money providers to run technology-based abnormal-transaction detection. Under the BOT's minimum practices for digital fraud management (ธปท.ว. 5713/2569), domestic systemically important banks and the e-money providers the BOT designates as significant must develop and start using, within 2026, an abnormal-transaction detection system that applies technology and can rank the risk of each transaction, account or customer; the practices name artificial intelligence and machine learning as examples of such technology. Other banks and e-money providers that judge such technology necessary must have a clear plan within 2026, and newly designated firms have one year from designation.
Sep 14, 2026GuidanceBOT Minimum Practice for Digital Fraud Management — Minimum Practice for Digital Fraud Management (แนวปฏิบัติขั้นต่ำสำหรับการบริหารจัดการภัยทุจริตดิจิทัล). On 14 September 2026 the Bank of Thailand issued minimum practices for digital fraud management by circular ธปท.ว. source ↗
Sep 12, 2025GuidanceBOT AI Risk Management Policy Guideline — Policy Guideline on Risk Management of the Use of Artificial Intelligence Systems (แนวนโยบายธนาคารแห่งประเทศไทย เรื่อง การบริหารจัดการความเสี่ยงของการใช้งานระบบปัญญาประดิษฐ์). The Bank of Thailand issued its policy guideline on risk management of the use of artificial intelligence systems on 12 September 2025, by circular ธปท.ว. source ↗
Jun 12, 2025ConsultationBOT Draft AI Risk Management Guideline — Draft Policy Guideline on Risk Management of the Use of Artificial Intelligence Systems (Guiding Principles for Artificial Intelligence Risk Management). From 12 to 30 June 2025 the Bank of Thailand consulted on a draft policy guideline on risk management of the use of artificial intelligence systems, which it titled in English Guiding Principles for Artificial Intelligence Risk Management. source ↗
  • Whether domestic systemically important banks and significant e-money providers have technology-based abnormal-transaction detection, able to rank the risk of each transaction, account or customer, in use by the end of 2026 as the September 2026 minimum practices require, and the abnormal-transaction patterns the BOT says it will specify for mule-account detection.
  • Any revision of the September 2025 AI guideline, an official English version, or further BOT guidance on generative or agentic AI; the guideline and its Q&A were published in Thai.
  • How BOT supervisors apply the AI guideline when examining banks' AI governance, data and model controls, given that it is framed as guidance applied by use case rather than as a notification with penalties.

Does the Bank of Thailand have AI rules for banks?

Yes, in the form of guidance. On 12 September 2025 the BOT issued a policy guideline on risk management of the use of artificial intelligence systems (circular ธปท.ว. 5994/2568) to commercial banks, specialized financial institutions and non-bank payment service providers. It covers board-level AI governance under responsible-AI principles such as FEAT and controls on data, models and AI-specific cyber threats. It is published in Thai.

Is the Bank of Thailand AI guideline mandatory?

It is a policy guideline, not a notification with penalties. Its practices are framed as what providers should do, and the BOT's appended Q&A says providers apply its principles and approaches according to how they use AI, while still complying strictly with laws such as personal data protection and intellectual property law. The BOT's separate digital-fraud minimum practices of September 2026, which name AI and machine learning, are mandatory.

Does the BOT AI guideline cover generative AI and third-party AI tools?

Yes. It asks providers using generative AI to reduce the risk of false outputs, for example through retrieval-augmented generation or prompt engineering, and to set a data boundary that separates corporate data when they use a third party's AI. Its Q&A says providers can apply it whether they build an AI system themselves or use an off-the-shelf system such as a generative AI chatbot.

Does the Bank of Thailand require a human in the loop for AI decisions?

Not as a blanket rule. For AI used in core functions tied to strategic decisions, such as approving loans, account openings, deposits, withdrawals or transfers, or in work that directly affects customers, providers may consider human-in-the-loop or human-over-the-loop oversight as appropriate, setting the level of human involvement by the risk and impact of the use.

Follow every move these regulators make

when one of these regulators moves, the next morning's brief says so · six sourced stories · 7 am PT · free

plus every regulation, bank and agent page update, the morning after · leave any morning