State Financial Regulators Bring AI Into the Examination Process

AI bank testing

State financial regulators are moving artificial intelligence oversight into routine supervision, giving banks and nonbank financial institutions more concrete expectations for documenting how they use the technology, manage its risks and protect consumers.

    Get the Full Story

    Complete the form to unlock this article and enjoy unlimited free access to all PYMNTS content — no additional logins required.

    Subscribe to our daily newsletter, PYMNTS Today.

    By completing this form, you agree to receive marketing communications from PYMNTS and to the sharing of your information with our sponsor, if applicable, in accordance with our Privacy Policy and Terms and Conditions.

    A new Artificial Intelligence Supervisory Framework from the Conference of State Bank Supervisors provides a model for those examinations. Alongside guidance from Michigan, Connecticut and New York, it signals that institutions should prepare to explain their AI practices to examiners, according to an advisory by law firm Reinhart Boerner Van Deuren.

    The developments address different aspects of AI oversight, from consumer outcomes and management accountability to cybersecurity. Together, they point toward greater scrutiny of AI across financial institutions’ products, services and internal operations.

    Released in September, the CSBS framework gives state regulators a standardized approach to examining AI use at regulated banks and nonbank institutions. It draws on existing risk management resources, including federal and industry AI frameworks, rather than developing its examination approach in isolation.

    The framework combines a core supervisory approach with a detailed, step-by-step examination guide. It includes sample initial scoping questions, document request lists, examination procedures, illustrative use cases and associated risk profiles. It also provides specific questions addressing generative AI and other emerging applications.

    CSBS supplemented the framework with a risk-tiering guide to help examiners identify high-risk uses, as well as specific guidance for agentic AI and third-party AI tools. Those additions place particular attention on institutions’ reliance on outside technology and artificial intelligence systems capable of taking actions.

    Reinhart expects the framework to become a guiding model for many states and shape future examinations. The immediate preparation step is an inventory: institutions need to identify where AI appears in their products, services and operations. Those using agentic AI or third-party tools should pay particular attention because the framework singles them out for heightened scrutiny.

    Michigan’s Department of Insurance and Financial Services issued a January bulletin addressing AI systems used by financial service providers. According to Reinhart, the department expects institutions’ decisions, including those concerning AI systems and applications, to be neither inaccurate nor arbitrary, capricious or unfairly discriminatory.

    Michigan also expects regulated entities to maintain a written AI program covering development and use, with systems designed to mitigate the risk of “Adverse Consumer Outcomes.” The bulletin reaches beyond institutions that formally deploy artificial intelligence: even entities deciding against its use are encouraged to establish, at minimum, an employee acceptable-use policy.

    Connecticut’s Department of Banking took a more explicit examination-focused step in September with guidance on AI systems governance. It applies to all entities the department licenses or regulates.

    The department confirmed plans to review AI systems and their governance frameworks, including senior management involvement, data security and implementation of AI governance plans. Reinhart advises affected institutions to develop and implement the frameworks Connecticut now expects before future supervisory examinations.

    New York’s September guidance approaches AI through cybersecurity. The Department of Financial Services’ guidance on risk assessments required by its cybersecurity regulation advises covered entities to consider whether adopting or using new technologies, including AI, materially changes their risk profiles.

    For institutions operating in New York, that means reviewing how new AI tools affect cybersecurity exposure, particularly risks involving sensitive data and information.

    Reinhart recommends developing, documenting and implementing artificial intelligence policies and practices; establishing procedures governing internal use and third-party vendor tools; and assessing and monitoring each application’s evolving capabilities and risks. Institutions should also track emerging state and federal guidance.

    The practical implication is that approving an AI tool is only the beginning of the governance process. As state examination expectations develop, institutions will need documentation showing how they oversee its use and respond to changing risks.