Correlated financial fragility

A market works in part because those taking part get things wrong in different ways. The risk here is not that a model fails, but that many institutions decide with the same model, the same data and the same provider, and stop being independent decisions.

The institutional map is unusually consistent. In November 2024 the Financial Stability Board named four vulnerabilities with systemic potential: third-party provider concentration, market correlation from the use of the same models and data, cyber risk from AI adoption by malicious actors, and model risk from opacity ^[430]The Financial Stability Implications of Artificial IntelligenceFinancial Stability Board · 2024 · reportView source ↗Accessed on 9 September 2026. A year later it measured the concentration with Bank of England figures: 33% of use cases are implemented by third parties —17% in 2022— and the three main model providers account for 44% of the providers named, against 18% two years earlier ^[431]Monitoring Adoption of Artificial Intelligence and Related Vulnerabilities in the Financial SectorFinancial Stability Board · 2025 · reportView source ↗Accessed on 9 September 2026. The mechanism is formulated by the BIS: institutions using the same algorithms could amplify procyclicality and volatility by exacerbating herding behaviour, liquidity hoarding, runs and fire sales ^[144]Artificial intelligence and the economy: implications for central banks (Annual Economic Report 2024, capítulo III)Bank for International Settlements · 2024 · reportView source ↗Accessed on 9 September 2026. On 25 June 2026 the European Systemic Risk Board raised the register: not a report but a formal warning under Article 16, published in the Official Journal ^[381]Warning of the European Systemic Risk Board of 25 June 2026 on systemic cyber risks stemming from frontier artificial intelligence models (ESRB/2026/3)European Systemic Risk Board · 2026 · official documentView source ↗Accessed on 9 September 2026.

What this does not demonstrate. None of the four vulnerabilities has shown up in a documented market episode: this is prudential surveillance, not a diagnosis of harm that has occurred. The Bank of England and FCA survey bounds the imminence: 75% of firms already use AI, but only 2% of use cases have fully autonomous decision-making ^[155]Artificial intelligence in UK financial services - 2024Bank of England; Financial Conduct Authority · 2024 · surveyView source ↗Accessed on 9 September 2026. And the FSB itself states that it cannot properly measure what it is watching: it is up against the speed of change and the lack of data on AI use in the sector ^[430]The Financial Stability Implications of Artificial IntelligenceFinancial Stability Board · 2024 · reportView source ↗Accessed on 9 September 2026 —which prevents both asserting the risk and ruling it out.