BankingNewsAI Daily Brief · Thursday, April 23, 2026
Banking AI
Financial institutions & fintech technology
US regulators just raised the bar for AI/analytics models at large banks (and exams will follow)
The OCC, Federal Reserve, and FDIC issued revised model risk management guidance for large banking organizations—material because it resets supervisory expectations for how banks validate, monitor, and govern models, including fast-proliferating genAI and agentic tooling. The update tightens end-to-end lifecycle controls (inventorying, testing, change management, ongoing monitoring) and will become the de facto yardstick for exams and MRAs when models touch credit, fraud, ALM, compliance, or customer outcomes.
Action
Accelerate a gap assessment against the revised guidance across your model inventory (including “non-traditional” genAI/LLM components, vendor models, and embedded models in platforms). Prioritize near-term remediation on documentation, independent testing, drift monitoring, and third‑party model governance so you can defend decisions under examiner scrutiny.
Citi and Google put an AI ‘team member’ in front of wealth clients—this is no longer just advisor tooling
Citi Wealth launched “Citi Sky,” an AI-powered assistant built with Google Cloud and DeepMind technologies, initially for US Citigold clients. The notable change is distribution: the assistant is positioned as an always-on member of the wealth team, pushing genAI from internal productivity into a client-facing channel where suitability, disclosures, recordkeeping, and complaint risk become first-order concerns.
Action
Treat client-facing genAI as a regulated channel: lock down advice boundaries, implement supervision/archiving, and run red-team tests for hallucinations and mis-selling scenarios before scaling. Benchmark your own wealth roadmap against Citi’s move—clients will now expect 24/7 AI service, not just human availability.