AI and Regtech

IMF News, October 29, 2021

Source details

Canonical URL

AI and Regtech

Other formats

• Markdown version
• Structured JSON version
• Bundle manifest

Bibliographic details

• Published: October 29, 2021

Overview

• Speech by Tobias Adrian, Financial Counsellor and Director, Monetary and Capital Markets Department, IMF.
• Event: Virtual Workshop on AI&Finance, October 29, 2021.
• Core message: AI is a technology with potential for enormous societal and economic impact, bringing in new opportunities and benefits, while creating new concerns and actions needed to mitigate potential risks.

Opportunities and benefits of AI in finance

• Competitive pressures are fueling wide adoption of AI in the financial sector.
• AI facilitates:
  • gains in efficiency and cost savings,
  • reshaping client interfaces,
  • enhancing forecasting accuracy,
  • improving risk management and compliance,
  • strengthening prudential oversight and equipping central banks with new tools for monetary and macroprudential mandates.
• COVID-19 pandemic accelerated adoption:
  • Banks increasingly used AI to handle high volume of loan applications, improve underwriting, detect fraud, and comply with mandated COVID-19 relief requirements.
• Potential future: emergence of closed loop AI-driven financial ecosystems where credit, risk management, compliance (regtech), oversight (suptech), and remedial actions are made by AI applications communicating directly.

Regtech and suptech use cases and impacts

• Regtech applications cut across banking, securities, insurance, and other financial services, including:
  • identity verification,
  • AML/CFT,
  • fraud detection,
  • risk management,
  • stress testing,
  • micro-prudential reporting,
  • macro-prudential reporting,
  • compliance with COVID-19 relief requirements.
• Specific regtech impacts:
  • AI reduces false positives in AML/CFT alerts, allowing reallocation of resources to likely suspect cases.
  • AI-enabled data analytics improve analysis of complex balance sheets and stress testing to meet regulatory requirements.
  • Emerging applications: mapping and updating regulatory obligations, conduct risk management (e.g., monitoring sales calls to ensure accurate disclosure of product features and risks).
• Suptech potential:
  • AI in microprudential supervision could improve assessment of credit and liquidity risks, governance and risk culture, enabling supervisors to focus on risk analysis and forward-looking assessments.
  • AI applied to market surveillance to detect collusive behavior and price manipulation.

Suptech examples

• ECB: AI-enabled machine reading of the “fit and proper” questionnaire to flag problems; search in supervisory review decisions to identify emerging trends and clusters of risks.
• Bank of Thailand: AI analysis of board meeting minutes to assess board regulatory compliance.
• De Nederlandsche Bank: AI data-analytics to detect networks of related entities to assess exposure to networks of suspicious transactions.
• Banca d’Italia: exploring AI in loan default forecasting.
• Monetary Authority of Singapore: project using AI algorithms for credit risk assessments by supervisors.

Risks, challenges, and robustness concerns

• General concerns with AI in finance:
  • embedded bias in AI algorithms,
  • opaqueness of outcomes,
  • robustness to cyber threats and privacy issues,
  • new sources and transmission channels of systemic risks (greater homogeneity in risk assessments and rising interconnectedness that could quickly amplify shocks).
• Suptech-specific challenges:
  • effectiveness depends on data standardization, quality, and completeness, particularly when leveraging non-traditional data sources;
  • resource and skills gaps require expanding personnel pool to include AI specialists and data scientists;
  • supervisors face risks similar to supervised institutions: privacy, cyber risk, outcome explainability, and embedded bias risk.
• Maintaining public trust in the robustness of AI algorithms is fundamental to financial stability.
• Robustness concerns include:
  • minimizing false signals during periods of structural shifts,
  • governance over AI development processes to strengthen prudential oversight and avoid unintended consequences.

Cyber risks and privacy (including data-manipulation attack types)

• AI introduces unique cyber risks and privacy issues beyond traditional cyber threats.
• Novel threats: manipulation of data at stages of the AI lifecycle to evade detection, cause wrong decisions, or illicitly extract information.[1]
• Privacy concerns:
  • AI may unmask anonymized data,
  • AI may “remember” information about individuals in the training set,
  • algorithm outcomes may “leak” sensitive data directly or by inference.
• Data manipulations include:
  • (a) data poisoning attacks that influence AI algorithm during the training stage and cause the AI to incorrectly “learn” to classify or recognize information, or create trojan models, which hide malicious actions that wait for special inputs to be activated, and
  • (b) Input attacks that allow attackers to introduce perturbations to data inputs and mislead AI systems during operations.

Policy recommendations and priorities

• Regulators should broadly welcome AI advancements in finance while preparing to capture benefits and mitigate risks—consistent with the Bali Fintech Agenda’s call to embrace the fintech revolution.
• Key policy actions:
  • develop clear minimum standards and guidelines for the sector;
  • strengthen focus on securing necessary technical skills for authorities and supervised institutions;
  • promote cooperation and knowledge-sharing at national, regional, and international levels to coordinate legal and regulatory actions supporting safe AI deployment;
  • foster collaboration among financial institutions, central banks, financial supervisors, and other stakeholders to counter interconnected risks.
• Monitoring and prudential oversight need strengthening because the evolving nature of AI means users, developers, and regulators do not yet fully understand the technology’s strengths and weaknesses; unexpected pitfalls may still materialize.

Source: Speech "AI and Regtech" by Tobias Adrian, Virtual Workshop on AI&Finance, October 29, 2021.

References

• Tobias Adrian
• People's Republic of China - Hong Kong Special Administrative Region and the IMF
• Speeches
• PRESS CENTER
• [Powering the Digital Economy: Opportunities and Risks of Artificial

Intelligence in Finance](https://www.imf.org/en/Publications/Departmental-Papers-Policy-Papers/Issues/2021/10/21/Powering-the-Digital-Economy-Opportunities-and-Risks-of-Artificial-Intelligence-in-Finance-494717)

• https://www.imf.org/en/home