Regulations for AI use in banks on the cards

KARACHI: The State Bank of Pakistan (SBP) is in the advanced stages of finalising guidelines for the responsible use of artificial intelligence (AI) in financial services.

These guidelines aim to promote trust, transparency, and accountability in AI-driven financial services while protecting the rights and interests of consumers.

In its Financial Stability Review 2024, the central bank noted that with AI adoption surging across different sectors, the global banking industry is also increasingly embracing these technologies.

In the case of Pakistan, financial institutions are also embracing AI technology. The most commonly adopted AI technologies by banks include robotics, process automation (for structured operational tasks), virtual assistant (for customer service), and machine learning techniques (for fraud detection and risk management).

SBP survey finds half of regulated entities have adopted AI tools

In 2024, the SBP surveyed 55 regulated entities (REs) to explore the current state, challenges, and future prospects of AI adoption within Pakistan’s financial services sector. The surveyed REs included commercial banks (conventional and Islamic), microfinance banks, digital banks, Electronic Money Institutions (EMIs) and Payment Service Operators/Payment Service Providers (PSOs/PSPs).

The results of the SBP’s survey on using AI in banks reveal that about half of the REs have either deployed AI in financial services or are in the development process.

Moreover, the survey indicates that AI is being used for various financial services, including fraud detection, customer services, marketing, credit risk assessment, process automation, etc. In this context, banks need to take care of the associated risks of AIs and incorporate the carbon footprint of AI into their risk management frameworks, introducing it as a specific risk category.

This requires relevant risk identification, assessment, development and implementation of adequate risk mitigation strategies.

“The foremost step is to recognise the potential environmental risks associated with adopting AI systems,” said the report.

The most commonly used AI technologies in the banking sector are among the most energy-intensive owing to real-time processing and high accuracy requirements.

“After risk identification, banks need to measure the emissions associated with AI models throughout their life cycle,” said the report.

“Banks must implement risk mitigation strategies to contain the environmental impact of their AI systems. The energy-efficient AI models and algorithms could help address climate concerns,” said the report.

Notably, International Financing Reporting Standards (IFRS) S1 and S2 also require companies to disclose information about all sustainability-related risks and opportunities, specifically about climate-related exposures.

“In Pakistan, the adoption of these standards have been proposed in a phased manner, initially starting with listed companies based on certain criteria, such as total assets, turnover and number of employees,” said the report.

It deserves emphasis that banks must define risk appetite and tolerance level of AI’s carbon footprint, the report added. By setting thresholds, banks would ably deploy AI system balancing operational efficiency with environmental concerns, it added.

The report said that a highly important aspect of AI is to assess the systemic risks on financial institutions and their customers. For instance, high technological penetration and supplier concentration could expose financial institutions to losses arising from operational failures, cyber-attacks and supply chain disruptions affecting key vendors. Further, increased use of AI models with similar characteristics could augment asset price vulnerabilities due to increased correlations in financial markets.

Published in Dawn, April 27th, 2025