Artificial intelligence is moving rapidly from experimentation to enterprise-wide adoption. Across banking, insurance, financial services and other industries, organisations are using artificial intelligence to assess creditworthiness, detect fraud, automate customer service, improve underwriting, monitor transactions and support strategic decisions. These capabilities offer substantial benefits in speed, efficiency and analytical depth. However, as artificial intelligence becomes embedded in critical processes, it also creates risks that cannot be addressed through technology controls alone.
The central challenge is no longer whether organisations should adopt artificial intelligence. It is whether they can govern it responsibly.
The Financial Stability Board’s recent consultation on sound practices for responsible artificial intelligence adoption by financial institutions reflects this changing priority. The proposed practices encourage boards and senior management to consider artificial intelligence within business strategy, technology adoption and risk management, rather than treating it as a standalone technical initiative. The guidance is not intended to prescribe a single model of adoption, but it reinforces the need for clear safeguards, proportionate governance and accountability at the highest levels.
Artificial intelligence risk extends well beyond model accuracy. Poor-quality or incomplete data can produce unreliable decisions. Opaque systems may make it difficult to explain why a customer was denied credit, flagged for investigation or charged a particular premium. Bias may become embedded in automated outcomes. Generative artificial intelligence can produce inaccurate or fabricated information with considerable confidence. Cybercriminals can use the same technology to conduct sophisticated fraud, impersonation and social-engineering attacks.
Institutions must also consider their growing dependence on external technology providers. Many artificial intelligence systems rely on third-party models, cloud infrastructure, specialised chips, data vendors and application programming interfaces. Concentration among a relatively small number of providers may create common points of failure across the financial system. The Financial Stability Board has identified third-party dependencies, market correlations, cyber risk and weaknesses in model governance as vulnerabilities that may be amplified by wider artificial intelligence adoption.
Responsible governance begins with clarity. Organisations need a complete inventory of artificial intelligence applications, clearly defined ownership and an understanding of which systems influence material decisions. Higher-risk applications should be subject to stronger approval, testing, validation and monitoring requirements. Human oversight must remain meaningful, particularly where decisions affect customers, employees, investors or regulatory obligations.
Accountability cannot be transferred to a software provider. Even when an artificial intelligence solution is purchased from a vendor, the organisation using it remains responsible for the decisions, risks and consequences arising from its use. Boards and senior management must therefore understand not only the benefits of artificial intelligence but also its limitations, dependencies and potential failure modes.
A responsible framework should establish acceptable-use boundaries, data-governance standards, independent challenge, cybersecurity safeguards, incident escalation and procedures for overriding or suspending a system where necessary. It should also address fairness, explainability, privacy, intellectual-property exposure and the use of confidential information in external artificial intelligence tools.
Governance must remain dynamic. Artificial intelligence systems can change as data, user behaviour and external conditions evolve. A model that performs well at deployment may become unreliable over time. Continuous monitoring, periodic reassessment and clearly defined performance thresholds are therefore essential. Institutions must also prepare for incidents in which an artificial intelligence system produces harmful, discriminatory or materially incorrect outcomes.
Strong governance should not be viewed as an obstacle to innovation. On the contrary, it creates the confidence required to adopt artificial intelligence at scale. Organisations that understand their systems, control their risks and retain effective human oversight will be better positioned to realise the benefits of innovation without compromising trust.
The speed of artificial intelligence development will continue to challenge traditional governance cycles. Policies reviewed annually may be insufficient for technology that evolves in months or even weeks. Risk, technology, compliance, legal, audit and business teams must work together continuously rather than operating in separate silos.
Artificial intelligence may transform how organisations make decisions, but responsibility for those decisions must remain human and institutional. The organisations that succeed will not necessarily be those that adopt artificial intelligence first. They will be those that combine innovation with accountability, transparency and sound judgement.
In the emerging artificial intelligence economy, responsible governance will not merely protect organisations from downside risk. It will become a foundation for customer confidence, regulatory credibility and sustainable competitive advantage.
Authored By
Dr. Rakesh Agarwal
Secretary General,
Risk Management Association of India (RMAI)
