The Basel Committee on Banking Supervision (BCBS) met on 28-29 September 2026 in Indonesia and formally "exchanged views on AI developments" as part of its work programme. This marks the first time AI has appeared as an explicit agenda item in BCBS's published meeting record — and it changes the regulatory calculus for every bank that operates under a BCBS member jurisdiction, including India.
No standard, consultation paper, or guidance document has been published. The September meeting was an information exchange. But the work programme entry is significant: it signals that a formal BCBS product on AI is in development. For Indian banks, already operating under RBI's MRMF and FREE-AI frameworks, the question is not whether BCBS will publish an AI standard — it is when, what it will cover, and how well their current compliance position will map to it.
What Happened at the September 2026 BCBS Meeting
The Basel Committee on Banking Supervision met on 28-29 September 2026 in Bali, Indonesia. Among the items formally recorded in the BCBS work programme: the Committee "exchanged views on AI developments in the banking sector."
This is significant for what it is and is not. It is not a consultation paper, a set of principles, or a standard. It is the first formal acknowledgment in BCBS's published record that AI governance is an active work programme item — the institutional equivalent of putting AI on the Basel agenda.
BCBS governs international banking standards: Basel III capital adequacy, operational risk, model risk management, and supervisory coordination. Its guidance shapes what central banks — including RBI — adopt into domestic frameworks. When BCBS publishes an AI standard, it becomes the reference point for bank supervision globally. RBI, as a full BCBS member, will translate it into India-specific guidance.
The September 2026 meeting was an information exchange, not a rulemaking event. No paper, no consultation, no standard. The significance is in the work programme entry — BCBS has formally started the clock on AI governance as an international banking standard.
Why This Matters for India
India is the incumbent G20 presidency holder for 2023 and a prominent G20 voice on digital finance. Three India-specific implications follow from the BCBS AI work programme entry.
G20-BCBS Alignment
India's RBI has been an active voice in G20 financial regulatory discussions. A BCBS AI standard will be shaped by G20 member input including India. RBI has published more detailed AI governance frameworks — FREE-AI (August 2025, 26 recommendations) and MRMF (June 2026) — than most G20 central banks. India is positioned to influence the BCBS standard rather than only adopt it.
The MRMF's NIST AI RMF alignment and its explicit provisions on GenAI (hallucination controls, red-teaming, kill switch, prompt injection) already address terrain that BCBS will need to cover. Indian regulators entering BCBS AI working groups bring a live, deployed framework as a reference point.
Pre-Emptive Compliance
Indian banks that already comply with RBI MRMF and FREE-AI will be substantially ahead of the BCBS curve when a standard issues. The convergence architecture between MRMF and global frameworks — NIST AI RMF, FSB 12 Sound Practices (June 2026) — means MRMF-compliant banks will have minimal gap to a future BCBS standard.
This is a structural competitive advantage. Banks in jurisdictions without detailed AI governance frameworks will face significant remediation when BCBS publishes. Indian banks that have built MRMF compliance can demonstrate readiness to international counterparties, correspondent banks, and foreign investors.
Examination Timing Window
BCBS typically moves from work programme entry to consultative document in 12-18 months. A BCBS AI consultation paper in Q1-Q2 2028 is plausible. Indian banks have a 2026-2027 window to build MRMF-aligned governance that will map to the BCBS standard when it arrives — rather than scrambling to retrofit compliance after the standard publishes.
What BCBS Will Likely Address
Based on FSB's 12 Sound Practices (June 2026), G20 AI governance discussions, and BCBS's own model risk history (the influence of SR 11-7 on global model risk frameworks), the first BCBS AI standard is likely to cover five domains:
| Domain | Expected BCBS coverage | India MRMF status |
|---|---|---|
| AI model risk governance | Extending SR 11-7 / domestic model risk concepts to a global standard covering LLMs and GenAI | Covered — MRMF's model inventory, validation independence, and documentation requirements directly address this |
| AI concentration risk | Risk from shared LLM API vendors — FSB flagged 88% of LLM API spend concentrated in 3 providers; BCBS likely to address systemic risk from AI vendor concentration | Partially covered — FREE-AI and MRMF address third-party AI vendor governance; concentration risk at systemic level is less explicit |
| AI in capital models | Use of AI/ML in internal ratings-based (IRB) models, stress testing, and ICAAP — high-stakes use cases already under Basel framework | Covered — MRMF applies to all model risk including credit risk models; MRMF Tier 1 classification captures IRB-equivalent AI models |
| Agentic AI governance | FSB explicitly noted the governance gap for autonomous AI agents; BCBS likely to address authorization boundaries, audit trails, and oversight requirements for agentic systems | Gap — MRMF has zero provisions for agentic AI. This is the primary area where Indian banks will need to build ahead of any BCBS standard |
| Operational resilience for AI | AI-specific recovery and continuity requirements; BCP for AI-dependent processes; AI system availability standards | Partially covered — RBI IT Governance Master Direction (November 2023) covers BCP/DR for AI infrastructure; MRMF adds kill switch requirements |
How BCBS AI Will Interact With RBI MRMF
RBI MRMF (June 2026) is already NIST AI RMF-aligned and covers the terrain BCBS will likely address. The convergence between MRMF and the anticipated BCBS standard is likely to be high — which means Indian banks building MRMF compliance now are simultaneously building toward BCBS readiness.
Key alignment points:
- Model inventory: MRMF's eleven-entity scope and model documentation requirements will map closely to any BCBS model risk extension. The inventory requirement is foundational across all frameworks.
- Validation independence: MRMF's independent validation requirements already exceed what most Basel member banks have implemented. BCBS will likely set a minimum standard that MRMF-compliant banks already meet.
- GenAI governance: MRMF's six explicit GenAI provisions — hallucination controls, mandatory red-teaming, tested kill switch, prompt injection controls, foundation model documentation, human escalation — are ahead of any BCBS member country except India. When BCBS addresses GenAI governance, MRMF will be a natural reference point.
- Third-party AI: MRMF and FREE-AI's vendor accountability provisions mirror FSB's concentration risk concern. Indian banks that have implemented vendor governance under MRMF will have a strong foundation for any BCBS third-party AI standard.
The gap that matters: MRMF has no agentic AI provisions. A BCBS standard will almost certainly address this — FSB's June 2026 Sound Practices explicitly noted the agentic AI governance gap, and BCBS follows FSB's analytical lead. Indian banks should build agentic AI governance frameworks now rather than waiting for BCBS to mandate it.
The India BFSI First-Mover Position
India's AI governance framework stack — RBI FREE-AI (August 2025) plus MRMF (June 2026) — is the most complete banking AI governance framework globally. A comparison with other BCBS member jurisdictions as of October 2026:
| Jurisdiction | Current AI governance framework | BCBS readiness |
|---|---|---|
| India (RBI) | FREE-AI August 2025 (26 recommendations, 6 pillars) + MRMF June 2026 (NIST-aligned, GenAI provisions, eleven-entity scope) | Highest — most complete framework; only gap is agentic AI |
| Singapore (MAS) | AI RMG consultation November 2025; Fairness, Ethics, Accountability, Transparency Toolkit March 2026 | High — advanced framework; closest to India's completeness |
| US (Fed/OCC) | SR 26-2 April 2026 (traditional model risk); agentic AI explicitly excluded from scope | Moderate — SR 26-2 covers model risk but excludes newer AI forms; significant gap on agentic AI |
| EU (ECB/EBA) | EU AI Act August 2026 enforcement (risk classification); no banking-specific model risk standard | Moderate — EU AI Act provides risk classification but no model risk framework equivalent to MRMF |
| UK (FCA/PRA) | FCA Consumer Duty approach; no model risk standard; PRA principles-based guidance | Lower — no comprehensive AI model risk framework; principles-based approach will require significant build to BCBS standard |
| Australia (APRA) | Operational resilience guidance; no AI-specific standard | Lower — no dedicated AI governance framework; operationally-focused guidance only |
Indian banks that achieve MRMF compliance are in the strongest position when BCBS issues its standard. That compliance position is a competitive advantage in global capital markets — it signals regulatory maturity to international counterparties at a time when AI governance is becoming a due diligence criterion for cross-border banking relationships.
MRMF is the world's most complete AI model risk framework for banking. Its one structural gap is agentic AI governance. BCBS will almost certainly address this in its standard — FSB already flagged it explicitly. Indian banks should build agentic AI governance (authorization boundaries, audit trails, multi-agent coordination controls, kill switch for agent pipelines) in 2026-2027, before BCBS mandates it.
Frequently Asked Questions
When will BCBS publish an AI standard?
No timeline has been announced. Based on BCBS's work programme pace (consultative documents typically follow initial discussion by 12-24 months), a consultation paper in 2027-2028 is plausible. The September 2026 meeting is the starting gun, not the finish line.
Does RBI participate in BCBS AI discussions?
Yes. RBI is a full BCBS member (India joined in 2009). RBI Governor Shaktikanta Das served on the BCBS Group of Governors and Heads of Supervision until his term ended December 2024. Current RBI representation continues at the committee level.
How should Indian banks track BCBS AI developments?
Monitor BIS.org (bis.org/bcbs) for consultative documents and working paper releases. BCBS typically publishes a work programme update in Q1 each year listing active projects. When an AI consultation paper issues, Indian banks should map it to RBI MRMF to identify any gaps — the convergence is likely to be high given MRMF's NIST-aligned structure.
Will a BCBS AI standard supersede RBI MRMF?
No. BCBS sets minimum international standards; domestic regulators implement them through national frameworks. RBI will translate any BCBS AI standard into India-specific guidance, likely building on MRMF rather than replacing it. The more likely scenario is MRMF revision post-BCBS standard, adding any elements not already covered.
More AI governance and compliance guides
FSB Frontier AI G20 Letter: What India Banks Must Do
FSB Chair's August 2026 letter identified frontier AI as a financial stability risk. BCBS met October 1, 2026. What India banks must do before the RBI circular arrives.
Read →RBI MRMF June 2026: AI Model Risk Framework for India Banks
Complete analysis of RBI's Model Risk Management Framework — scope, tiers, GenAI provisions, and examination readiness.
Read →FSB 12 Sound Practices for AI in Financial Services 2026
FSB's June 2026 sound practices for AI governance — what they require and how India's frameworks compare.
Read →RBI AI Circular compliance checklist for Indian banks
Six obligation categories, what evidence an examiner will ask for, and common gaps.
Read →Preparing for BCBS AI compliance before the standard lands?
Zytra's India compliance team will map your current RBI MRMF coverage to anticipated BCBS AI requirements, identify gaps — especially on agentic AI — and design a compliance programme that satisfies both India and international standards simultaneously.