FRB: Speeches
Artificial Intelligence in the Financial System — Key Risks and Risk-Management Priorities
AI can improve efficiency and risk detection across financial firms but introduces material cyber, model, operational, concentration, and governance risks that require stronger risk management, vendor oversight, cross-sector coordination, and investment in resilience and expertise.
- AI presents tangible benefits for financial firms (automation, faster analytics, improved detection of fraud and illicit activity) alongside significant new risk vectors.
- Cybersecurity risk is amplified by AI: models can be targeted via data poisoning, adversarial attacks, and exploitation of automated decision pipelines, increasing potential for large-scale incidents.
- Model risk expands beyond traditional statistical issues to include opaque/complex models, distributional drift, brittleness, and reduced interpretability, complicating validation and auditability.
News
AIC minutes: systemic and governance risks from AI integration in UK financial services
Consortium members identified concentrated third‑party providers, model homogeneity, agentic AI, explainability gaps, data and talent shortages, and operational dependencies as primary risks from growing AI use in UK financial services, recommending standards, third‑party assurance, enhanced testing/evaluation, and evolving governance to manage contagion and edge‑case risks.
- Four focused workshops addressed: concentration in AI infrastructure, evolution of AI edge cases, explainability/transparency for generative AI, and AI‑accelerated contagion in markets.
- Concentration risk: rising reliance on a small number of third‑party AI providers creates single points of failure, contagion via model updates, capacity/scalability constraints, and calls for third‑party assurance and minimum standards.
- Contagion drivers identified: synchronized market dynamics from common vendors/models/data, operational resilience vulnerabilities from critical vendor dependencies, and model concentration/homogeneity causing correlated errors across firms.
News
Constraints to Responsible AI Adoption in UK Regulated Firms — Bank of England Roundtables
Regulated firms broadly welcome the PRA’s principles- and outcomes-based AI framework but identify practical constraints slowing responsible AI adoption: cautious second-line risk functions and legacy model-validation approaches, skills shortages, cross-jurisdiction regulatory fragmentation, procurement frictions with third‑party AI providers, data protection and sovereignty rules, and limited data quality in some insurance lines. Firms want evolved risk management focused on testing, monitoring and outcome guardrails, shared supervisory observations and international coordination.
- Broad support for the PRA’s principles- and outcomes-based regulatory approach and for Supervisory Statement 1/23 on Model Risk Management; most participants do not currently want AI-specific rules or a PRA/Bank AI sandbox.
- Second-line risk functions exercising heightened caution is delaying AI deployment; causes include limited AI skills/expertise and an emphasis on demonstrating compliance with supervisory expectations.
- Traditional model risk validation (focusing on understanding internal model mechanics) is increasingly unsustainable for generative and agentic systems; participants advocate shifting risk management toward robust testing, continuous monitoring and setting guardrails around system outcomes rather than full explainability.
News
AIC minutes: concentration, agentic edge cases, explainability and AI‑driven contagion risks in UK finance
Consortium workshops identify four priority risk domains: concentration of AI providers and infrastructure; novel high‑impact 'edge' uses driven by agentic AI; explainability and transparency for generative models; and AI‑accelerated contagion across markets. Cross‑cutting themes include third‑party dependencies, limits of human‑in‑the‑loop controls, need for targeted governance and scenario testing, and divergent adoption incentives across firm sizes.
- Concentration risk: heavy reliance on a small set of AI providers, foundation models, compute and specialist expertise creates correlated vulnerabilities and limits firms' visibility and control over model design, updates and resilience.
- Third‑party dependencies: short‑notice vendor updates and opaque supply chains (hardware, cloud, data, models) constrain firms' ability to assess risk, prompting collaboration with CMORG AI Taskforce to improve AI supply‑chain transparency.
- Agentic edge cases: emergent 'agentic' workflows that autonomously act across systems can produce novel, high‑impact operational risks and require tailored monitoring, governance and potentially system‑specific controls (e.g., predefined circuit breakers).
Capital Flows Research
Decomposing the Geopolitical Risk Premium Across Equities, FX and Rates
Markets are being driven by three distinct forces—AI thematic flows, geopolitical risk flows, and positioning unwinds—and quantifying each across equities, FX and rates reveals when and how they move markets; the report finds geopolitical risk has contributed roughly 30% of the S&P500's year-to-date move and argues for agentic macro models to identify timing and asymmetric trade opportunities.
- Three primary market drivers right now: AI thematic flows, geopolitical risk flows, and positioning unwinds; successful trading requires quantifying all three to map timing and impact.
- Agentic macro models are being built to quantify the macro regime and spotlight large, asymmetric bets aligned with regime dynamics.
- Geopolitical risk is a measurable premium that has materially impacted equities—estimated to explain ~30% of the S&P500/YTD move once persistent negative returns emerged.
Mind The Tape
AI in 2026: Year of the Agents — The Trade Broadens
AI agents are positioned to broaden trading activity and reshape market dynamics in 2026, creating new automation-driven flows and investment opportunities while forcing investors to adapt strategies and infrastructure exposure.
- Deployment of autonomous AI agents will expand the scope and volume of algorithmic trading and automated market activity.
- Market structure and participant behavior will shift as agent-driven flows create new liquidity patterns, execution demands, and arbitrage opportunities.
- Investment focus will move toward agent infrastructure, orchestration platforms, data and observability tools, and regulatory/operational risk management.
Capital Flows Research
Geopolitical Risk Premium: How Equities, FX and Rates Are Being Priced
Markets are being driven by three concurrent forces—AI thematic flows, geopolitical risk flows, and positioning unwinds—and quantifying each with agentic macro models reveals when and how they move equities, FX and rates. On a year-to-date basis roughly 30% of S&P500 moves are attributed to a geopolitical risk premium, with crude, short-term rates (STIR) and FX positioning unwinds amplifying equity volatility and creating divergence-driven opportunities.
- Three dominant drivers today: AI thematic flows, geopolitical risk flows, and positioning unwinds; understanding and quantifying all three is necessary to time market moves.
- The author has built agentic macro models intended to quantify the macro regime and identify large asymmetric trading opportunities aligned with those regimes.
- Approximately 30% of year-to-date S&P500 movement since persistent negative rolling returns in ES is attributed to a geopolitical risk premium.
Capital Flows Research
Inside the Machine: CFR's New Intelligence Suite and the Current Macro Regime
The author is building an AI-driven, agentic research and macro trading system to map the prevailing macro regime and identify a few large asymmetric trades; they argue AI converging with higher-quality data and expanding hardware infrastructure will precipitate a major market reckoning while markets underprice geopolitical risk.
- Developing an AI-powered, agentic research platform to drive macro trading decisions and generate asymmetric 'home run' trades.
- Belief that AI convergence with higher-quality data and accelerated hardware/robotics deployment will fundamentally shift competitive moats and create a brutal reckoning for unadapted players.
- Primary analytical focus is regime identification across growth, inflation, liquidity, and the credit cycle, with special emphasis on geopolitical risk that markets are mispricing.