The useful question is what changes for users, developers or buyers, and whether the announcement stays industry context or becomes something people can actually use.

The AI Consortium’s quarterly meeting, co-chaired by David Geale and Sarah Breeden, examined the rapid development of AI in financial services, emphasizing the need for regulators to support responsible adoption. Workshop discussions focused on explainability and transparency in generative AI, highlighting challenges such as system complexity, third-party opacity, and unpredictable outputs that complicate existing risk frameworks. Members proposed managing model risk from an AI system perspective, noting that interconnected models and independent updates require stronger governance and clearer boundaries to ensure proportional controls.

Workshop participants identified risks of AI-driven contagion in the financial system, including price volatility and operational failures due to automation and shared technical dependencies. They proposed a firm-level response taxonomy and stressed the need for cross-firm testing to detect both fast-moving and slow-moving contagion risks. Members also discussed the importance of defining agent action limits and reframing supervisory questions to assess system-wide safety rather than individual firm controls.

The consortium highlighted concentration risks in AI provision, particularly where services support critical business functions, and noted talent shortages in AI-related skills beyond engineering. Members suggested accelerating upskilling programs and strengthening domestic talent pipelines to address systemic gaps. They also explored whether model hosting (in-house vs. SaaS) affects dependency risks, with preliminary findings indicating greater reliance on SaaS models.

Workshop leads outlined a four-step approach to manage AI system failures, applicable across firms, and discussed the need for voluntary output reporting to enhance governance. Members reflected on recent advances in frontier AI, emphasizing the importance of cross-sector collaboration on cybersecurity and maintaining focus on ethics, bias, and system design to preserve trust and market stability.