What does BOT Draft AI Risk Management Guideline say about AI in banking?

Published Jun 12, 2025 · Last reviewed Nov 19, 2026

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. The draft set out board-level AI governance under FEAT principles (fairness, ethics, accountability, transparency) and data, model and cyber controls, and it proposed that customer-facing AI be disclosed before service with an option for customers to disable or bypass it. The BOT published a summary of the comments with its responses and issued the final guideline on 12 September 2025.

OFFICIAL TEXT: bot.or.th ↗ · SUPERSEDED · BANK OF THAILAND

Checked 2026-11-19 · Bank of Thailand, bot.or.th
DocumentBOT 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)
Issued byBank of Thailand (BOT), Thailand's central bank and the supervisor of commercial banks, specialized financial institutions and payment service providers
TypeConsultation
StatusSuperseded
PublishedJun 12, 2025
Comment deadlineJun 30, 2025
Applies toFinancial institutions, specialized financial institutions and regulated payment system and payment service providers, the parties the BOT consulted. A consultation draft, replaced by the final guideline of 12 September 2025. Published in Thai, with the English title Guiding Principles for Artificial Intelligence Risk Management
Superseded byBOT AI Risk Management Policy Guideline
Official sourcebot.or.th ↗
Use casesAI governance (general) · Customer-facing chatbots · Model risk management · Cybersecurity · Generative & agentic AI

What are the key points of BOT Draft AI Risk Management Guideline?

  • Process: one round of public comment on the BOT website from 12 to 30 June 2025, addressed to financial institutions, specialized financial institutions and regulated payment businesses.
  • Structure: governance (Part 1) and development and security (Part 2), the two parts the final guideline kept.
  • Human oversight in the draft: for AI used in strategic functions or customer interaction, providers should consider involving humans in decisions, choosing human-in-the-loop or human-over-the-loop by risk and impact; the draft's footnotes took both definitions from ETDA's AI Governance Guideline for Executives.
  • Customers in the draft: for customer-facing AI, providers should disclose AI use before providing the service, offer an option to disable or bypass AI-based services, and educate customers on safe use.
  • Risk tiers: respondents suggested classifying AI as high-risk or prohibited; the BOT kept its approach, saying the guideline sets principles and providers can grade the risk and impact of their AI under their own risk framework.
  • Other responses: comments argued that human oversight of every AI decision and a disable-or-bypass option for every AI service would strain resources and cut against the purpose of using AI, and the BOT reworded both so providers apply them according to risk; it clarified the AI definition and added a Q&A on excluded systems, data quality, test data, legal risks such as intellectual property and personal data, and off-the-shelf AI.

What did BOT Draft AI Risk Management Guideline change for banks?

The final guideline of 12 September 2025 replaced the draft. It kept the two-part structure and the FEAT reference, turned the disable-or-bypass option into informing customers when AI talks to them in place of staff and possibly offering a route to a human, made human-in-the-loop or human-over-the-loop oversight something providers may consider by risk for core functions and work that directly affects customers, and added a five-question Q&A.

What did the Bank of Thailand change after its AI consultation?

It softened the customer-facing clause, which had proposed letting customers disable or bypass AI-based services, into informing customers that AI is serving them and possibly offering a way to reach staff. It made human oversight a choice scaled to risk, clarified the definition of AI and added a Q&A on scope, data quality, testing, legal risk and third-party AI.

Did the Bank of Thailand add high-risk or prohibited AI categories?

No. Respondents suggested risk tiers such as high-risk or prohibited AI, but the BOT kept a principles-based guideline and said providers can grade the risk and impact of their AI systems under their own risk management framework.

DateDocumentStatus
Sep 12, 2025BOT AI Risk Management Policy Guideline — Policy Guideline on Risk Management of the Use of Artificial Intelligence Systems (แนวนโยบายธนาคารแห่งประเทศไทย เรื่อง การบริหารจัดการความเสี่ยงของการใช้งานระบบปัญญาประดิษฐ์)Final
Sep 14, 2026BOT Minimum Practice for Digital Fraud Management — Minimum Practice for Digital Fraud Management (แนวปฏิบัติขั้นต่ำสำหรับการบริหารจัดการภัยทุจริตดิจิทัล)In force
Nov 19, 2026BNM AI Feedback Statement — Feedback Statement: Artificial Intelligence (AI) in the Malaysian Financial SectorFinal
Nov 19, 2026BoE/PRA/FCA effective practices on AI cyber risk (Nov 2026) — Effective practices: Cyber risk and resilience management relating to artificial intelligence (AI)Final
Nov 19, 2026EBA Risk Assessment Report (Nov 2026) — Risk Assessment Report of the European Banking Authority, November 2026, with Box 1 on artificial intelligence related risks and opportunitiesFinal
Nov 17, 2026FSB Plenary: priorities for 2027 (Nov 2026) — FSB sets out its priorities for 2027 (Plenary meeting, Basel, 16-17 November 2026)Final

Follow every document these regulators publish

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