The Financial Stability Institute released a paper that canvasses a selection of policy documents on artificial intelligence (AI) governance issued by financial authorities or groups formed by them in nine jurisdictions and other cross-industry AI governance guidance that apply to the financial sector. It aims to provide a snapshot of existing regulatory approaches on AI governance, including initial thinking as expressed in consultation papers and to identify emerging common regulatory themes including from relevant cross-industry, general AI guidance.
The policy benchmarking exercise is supplemented by interviews with four financial authorities. Other regulatory authorities may draw inspiration from this stock take to consider adopting relevant regulatory approaches that are appropriate for their local circumstances.
At the international level, one of the main recommendations and principles that are commonly referenced in the financial sector are the Organisation for Economic Cooperation and Development (OECD) AI Principles and the G20 AI Principles – the latter drawing from the former’s recommendations. These principles cover five broad areas, namely: benefits to people and planet; respect for rule of law and human rights; transparency and responsible disclosure; continuous risk assessment and accountability.
AI-related guidance have also been issued at the regional or national levels (ie the EU and the United Kingdom respectively). Principles that specifically target the use of AI in the financial sector have been issued in the European Union, Germany, Hong Kong, the Netherlands, Singapore and United States. More recently, a proposed regulation laying down harmonized rules on AI across all industries, which is a first in the world, was issued in the EU.
The existing issuances revolve around five common principles – reliability/soundness, accountability, transparency, fairness and ethics. Other issues covered by the issuances include data privacy, thirdparty dependency and operational resilience. Most of these issues are the same ones authorities look at when assessing traditional models used by financial institutions.