GFMA submitted letter dated May 28, 2024, to leadership of FSB, Basel, IOSCO, CPMI, and OECD regarding key considerations for artificial intelligence (AI) in capital markets.
In particular, the letter includes a set of industry views concerning key considerations regarding the use of, and the regulatory approach to, AI in capital markets.
Document dated May 28, 2024, received from GFMA Jun. 5, summarized on Jun. 8.
Follows IOSCO May 2024 issued OECD-FSB roundtable on AI in finance, see #213388.
Follows EP May 2024 regulation setting out harmonized rules on AI, see #204700.
Definition of AI
GFMA does not endorse a specific definition of AI since it is neither a narrow nor static technology, instead favoring a principles-based and outcomes-focused approach.
However, for reference, GFMA adopts OECD definition of AI systems, which describes AI as machine-based systems inferring outputs from input to influence environments.
This definition is widely accepted, utilized by major jurisdictions, including EU's AI Act.
Existing Standards & Frameworks
Existing global standards and frameworks, such as those from the OECD, have proven effective and should be adapted rather than creating new AI-specific standards.
If gaps identified, updates to existing governance frameworks should be explored first, with new standards considered only if necessary, complementary to existing processes.
GFMA encourages global standard setters to ensure alignment, avoid unintended consequences; call on the G20 to endorse this approach, consistent with previous commitments, to promote international cooperation and governance for AI.
Model Risk Standards & Frameworks
Capital markets firms adhere to robust model risk standards and frameworks, which already encompass AI applications and models used by these firms.
These frameworks undergo regular risk-based and outcomes-based reviews.
Should any gaps in these frameworks be identified in the future, GFMA is open to collaborating with authorities to update them accordingly/provide necessary guidance.
Third-Party Risk Management Standards
Recently, organizations like FSB, IOSCO, Basel have adopted robust third-party risk management standards: technology-neutral, principles-based, outcomes-focused.
These standards are adaptable to the evolving landscape of AI, showcasing the effectiveness of such an approach in managing new AI use cases.
As global standard setters and regulators delve into third-party risks associated with AI, GFMA suggests considering all actors and responsibilities within AI supply chain.
Clearly defining the roles of all actors in the AI supply chain can help increase transparency and facilitate application of existing risk-based standards on this topic.
New AI-Specific Regulation
AI-specific regulation poses risk of stifling innovation in financial services, other sectors globally, if regulations inconsistent, unclear, overly prescriptive across jurisdictions.
Many jurisdictions have already enacted their own AI-specific regulations, leading to potential fragmentation and hindrance of technological advancement.
Instead of introducing new AI-specific standards, regulators should focus on applying existing technology-neutral and outcomes-based frameworks to AI use cases.
Recent initiatives have highlighted the importance of this approach.
International regulatory alignment is essential to establish cohesive AI governance standards that promote safe, secure, trustworthy, and sustainable AI.
Collaboration between official sector, industry encouraged to develop risk frameworks, toolkits, fostering innovation while mitigating unintended impacts on financial stability.