On 16 March 2022 the Basel Committee published a newsletter summarising its internal discussions on banks' use of artificial intelligence and machine learning. It states that AI/ML can raise operational efficiency and improve risk management but heightens risks around model complexity and interpretability, data quality and bias, cyber and third-party interconnection, and outsourcing accountability. The Committee named three areas for further work: explainability of outcomes, governance and accountability structures, and implications for bank resilience and financial stability.
| Document | BCBS AI/ML Newsletter (March 2022) — Newsletter on artificial intelligence and machine learning |
| Issued by | Basel Committee on Banking Supervision (BCBS) |
| Type | Guidance |
| Status | Final |
| Published | Mar 16, 2022 |
| Applies to | Informational for supervisors and internationally active banks; explicitly 'does not constitute new supervisory guidance' |
| Official source | bis.org ↗ |
| Use cases | Model risk management · AI governance (general) · Third-party & vendor AI |
What are the key points of BCBS AI/ML Newsletter (March 2022)?
- Issued 'for informational purposes only'; it is the Committee's first and still its only stand-alone AI/ML publication.
- Risk themes: complexity and interpretability of AI/ML models versus traditional approaches; data quality, relevance, security, confidentiality, bias, and ethics; cyber exposure from third-party interconnectivity and cloud.
- Banks remain fully responsible for models developed by third parties, including due diligence and oversight.
- Three priorities for further analysis: explainability and interpretability of outcomes; governance structures defining roles, responsibilities and accountability; implications for individual bank resilience and systemic financial stability.
- Stresses adequate skilled staff across developers, validators, users, and independent auditors.
- Referenced as the Committee's baseline in the May 2024 Digitalisation of finance report.
What did BCBS AI/ML Newsletter (March 2022) change for banks?
The newsletter created no obligations, but it set the vocabulary (explainability, governance, resilience) that national supervisors in the 28 Basel jurisdictions have since used in AI examinations, and it put on record that model risk management expectations extend to AI/ML and to vendor-built models.
Is the 2022 Basel AI/ML newsletter binding guidance?
No. The Committee states it does not constitute new supervisory guidance. It records supervisory observations that member authorities may apply through existing model-risk and operational-risk frameworks.
What did the Basel Committee say about explainability of AI models?
That AI/ML models are harder to interpret than traditional models, that transparency and explainability need attention, and that explainability of outcomes is one of three areas it would analyse further.
| Date | Document | Status |
|---|---|---|
| Jun 2, 2026 | BCBS ICT Risk Management Report (June 2026) — Information and communication technology risk management: range of practices | Final |
| Jan 6, 2026 | BCBS 239 Implementation Newsletter (Jan 2026) — Implementation of the Principles for effective risk data aggregation and risk reporting (BCBS 239 Principles) | Final |
| Dec 10, 2025 | BCBS Third-Party Risk Principles (Dec 2025) — Principles for the sound management of third-party risk | In force |
| Feb 4, 2025 | BCBS Work Programme 2025–26 — Basel Committee work programme and strategic priorities for 2025/26 | In force |
| May 16, 2024 | BCBS Digitalisation of finance report (May 2024) — Digitalisation of finance | Final |
| Mar 31, 2021 | BCBS Principles for Operational Resilience (2021) — Principles for Operational Resilience | In force |
Follow every document these regulators publish
6 curated AI stories for banking executives · Every morning · Free
Subscribe to BankingNewsAI →