Federal Reserve Vice Chair for Supervision Michelle Bowman stated this week that the Financial Stability Board's Standing Committee on Supervisory and Regulatory Cooperation is working with the U.S. Treasury and the SEC on a report covering sound practices for AI adoption, use, and innovation in financial services. A draft is expected in Q3 2026. The statement signals that federal regulators are moving from observation to framework-building on AI, and that the framework will arrive at roughly the same time banks are deploying their first production agentic systems.
The practical significance for equipment finance bank programs is direct. Any AI deployment inside an FDIC-supervised institution will eventually be subject to the guidance this report produces. Banks that are already running AI agents in credit, compliance, or fraud functions when the framework drops will face retroactive review. Banks that wait will face a cleaner implementation path but will have lost 12 to 18 months of operational learning.
The same week Bowman made her remarks, FIS announced a partnership with Anthropic to deploy an AI agent for financial crimes compliance combining Claude's reasoning with FIS's banking data and regulatory infrastructure. BMO and Amalgamated Bank are named as the first institutions to deploy the agent, with broader availability planned for the second half of 2026. Anthropic simultaneously announced 10 new AI agents targeting banks, insurers, and other financial services firms, including a connector to Moody's MCP platform.
The financial crimes use case is the lowest-risk entry point for agentic AI in banking. Regulators have been explicit that BSA/AML compliance is under-resourced at most institutions, and an agent that surfaces suspicious activity patterns from transaction data is easier to validate and audit than one making credit decisions. The FIS/Anthropic deployment is notable not because financial crimes AI is new, but because it is the first named production deployment of a Claude-based agent inside a major bank program, with two institution names attached to it.
The week also produced the second bank failure of 2026. The Georgia Department of Banking and Finance closed Community Bank and Trust - West Georgia in La Grange, Georgia on Friday. Total assets were $288 million and total deposits $268 million. The institution carried a FedFis Rating of 4.64 as of the December 2025 quarter, placing it in the lowest percentile of all banks. The FDIC has been assigned as receiver.
The juxtaposition is worth naming. In the same week that regulators announced a framework for AI in banking and major vendors announced agentic deployments at large institutions, a $288 million community bank failed with a risk rating that placed it in the bottom percentile. The institutions deploying AI agents and the institutions failing are not the same institutions. The technology investment gap between the top quartile and the bottom quartile of the banking system is widening, and the community bank failure rate, while still low in absolute terms, is a lagging indicator of that gap.