TL;DR

The FIX Trading Community has proposed 10 recommendations for regulating AI in trading, responding to a Singapore MAS consultation. The proposals call for a globally recognised AI taxonomy, board-level accountability, and specific safeguards for large language model integration in trading algorithms.

A Framework for AI in Capital Markets

The FIX Trading Community’s recommendations address the growing use of AI across trading operations, from algorithmic execution to risk management. The proposals aim to balance innovation with the stability requirements of financial markets, an environment where the consequences of system failures can be severe and immediate.

The 10 recommendations span several areas:

• Global standards: A globally recognised AI definition and taxonomy specific to trading, reducing fragmentation across jurisdictions.
• Governance: Board-level accountability for AI governance effectiveness, supported by cross-functional oversight committees that bring together technology, risk, and compliance expertise.
• Existing frameworks: Anchoring AI regulation to established resilience standards such as MiFID RTS 6 and DORA, rather than building entirely new regulatory structures.
• LLM-specific controls: Dedicated consideration of how large language models are integrated into trading algorithms, reflecting the distinct risks these models introduce.
• Transparency: A “food labelling” style disclosure approach, making it clear to market participants when and how AI is being used.
• Risk management: Expanded risk dimensions covering interconnectedness and contagion risks, as well as change sensitivity and nonlinearity in AI-driven systems.
• Lifecycle management: Strengthened controls across the full AI lifecycle, including revised definitions of what constitutes a “significant change” requiring additional oversight.

Capital Markets Need Different Rules

Executive Director Jim Kaye warned against importing Silicon Valley’s “move fast and break things” mentality into capital markets. The warning reflects a broader concern within financial services that the speed of AI adoption may be outpacing the development of appropriate controls and governance.

Looking Forward

The recommendations come as regulators worldwide grapple with how to oversee AI in financial markets without stifling innovation. The FIX community’s call for building on existing frameworks rather than creating entirely new ones may offer a pragmatic path forward, though the challenge of coordinating global standards remains considerable.