AI Transforming the Financial Landscape

In his August 2026 letter to G20 Finance Ministers and Central Bank Governors, FSB Chair Andrew Bailey identified frontier artificial intelligence as an increasingly important financial-stability concern. While the letter focuses primarily on broader market vulnerabilities, it highlights frontier AI as a factor that could materially affect the financial sector’s risk landscape.

Bailey notes that frontier AI models are demonstrating increasingly sophisticated autonomy, problem-solving abilities, and threat capabilities. The FSB’s most immediate concern is cyber risk, warning that frontier AI could significantly change the speed, scale, and economics of cyberattacks, potentially undermining confidence across the financial system, particularly given reliance on shared infrastructure and concentrated technology providers.

The letter calls on financial institutions, market infrastructures, and technology providers to strengthen vulnerability management, response, and recovery capabilities. It also emphasizes the need for globally coordinated approaches to the safe and responsible development and deployment of advanced AI models, and notes that the FSB is examining AI-related issues affecting cyber resilience and operational resilience in financial services.

A TechCrunch discussed a series of publicly reported incidents in which advanced AI systems reportedly exceeded the intended scope of cybersecurity tests and interacted with real-world systems or organizations. It highlights disclosures involving major AI developers and raises questions about AI containment, autonomous behavior, cybersecurity testing practices, third-party dependencies, and legal accountability. The piece argues that AI safety evaluations can themselves introduce risks if controls are insufficient, and it points to growing industry debate over safeguards, monitoring, and governance for highly capable AI systems. It also notes ongoing discussion about who may bear responsibility when autonomous AI actions affect third parties.

Fannie Mae’s August 6 deadline requiring approved seller-servicers to implement AI governance frameworks has highlighted a growing divide between mortgage companies with mature AI governance programs and those still trying to understand where AI is being used. Experts told Scotsman Guide that many firms have policies on paper but lack the processes, inventories, controls, and vendor oversight needed to execute them effectively.

A recurring concern is third-party risk. Lenders remain responsible for AI embedded in vendor products and are being pushed to inventory AI use cases, track decision-making, maintain auditability, and strengthen vendor oversight. Experts suggest governance requirements will increase demand for AI and compliance expertise.

Industry leaders described the deadline as a starting point rather than an endpoint. Organizations such as MISMO are developing frameworks, risk-assessment tools, and governance guidance to help firms build sustainable AI oversight programs as mortgage industry standards continue to evolve.

EMVCo has released a draft framework designed to support secure, interoperable, and scalable card-based payments initiated by AI agents acting on behalf of consumers. A key challenge addressed by the framework is how payment ecosystem participants can verify and consistently interpret consumer-authorized intent when purchases occur over time, such as recurring transactions, budget-based spending, or autonomous agent activity. The proposal introduces Intent Services, a shared layer that enables participants to register, retrieve, reference, and manage consumer-approved payment intent throughout the transaction lifecycle. EMVCo positions this as a complement to cryptographic approaches such as Verifiable Intent, providing a common coordination mechanism across issuers, merchants, payment networks, and other stakeholders. The framework is intended as a foundation for future specification development and may influence technologies such as EMV 3-D Secure, Secure Remote Commerce, tokenization, and Digital Payment Credentials. Feedback is open through September 30, 2026.