Board Meeting Digest — July 2025
AI in Credit Unions: The Guidance Gap
A GAO report issued in May 2025 found that credit unions may be at a supervisory disadvantage relative to banks because the NCUA lacks model risk management guidance comparable to the banking agencies' 2011 framework. The NCUA's response was deliberate: rather than issuing formal guidance that could become de facto requirements subject to examination, the agency is building an AI resources page linking to authoritative sources (NIST, CISA) and explicitly confirmed that credit unions can use AI tools without specific NCUA permission. The principle, as Chairman Hauptman framed it: "Just because NCUA doesn't have guidance on a specific tool or activity doesn't mean it's necessarily impermissible."
The use cases presented are concrete. One credit union reduced call volume by 40% using machine learning models for member behavior analysis. Others use AI-based check fraud detection to process significantly more items with lower losses. Foreign-language chatbots are expanding member access in multilingual communities. Board Member Harper pushed the agency on whether it could pre-vet AI vendors for smaller institutions that lack in-house expertise — staff acknowledged the merit but flagged the endorsement risk.
The broader signal: the regulatory posture toward AI in credit unions is permissive, but the absence of formal guidance creates uncertainty that may slow adoption, particularly among the Under $100M institutions that would benefit most from efficiency gains.
Credit union responsibility remains unchanged
All three board members emphasized a consistent point: AI does not change accountability. Chairman Hauptman: "You can't blame your AI for something. The credit union CEO is responsible and no other human on earth." Whether AI is used for lending decisions, member communications, or fraud detection, the credit union bears the same regulatory responsibility as if the task were performed manually. Consumer protection requirements, information security obligations, and fair lending compliance apply regardless of the underlying technology.
NCUA's own AI adoption
Internally, the NCUA is using AI for 5300 Call Report data quality review — machine learning algorithms identify reporting anomalies, saving examiners approximately 40 hours per quarter. The agency is also piloting AI for document creation, data analysis, and meeting summaries. Staff noted that even modest time savings across the workforce — "a few hours each month" — can justify investment, particularly given the 20%+ workforce reduction from the voluntary separation program.
Central Liquidity Facility: Membership Up 28%
Central Liquidity Facility membership grew from 350 to 447 credit unions — a 28% increase — with roughly half of members below $250 million in assets. The growth is notable in context: Q2 2025 produced the widest gap between loan growth (1.80% QoQ) and deposit growth (0.22% QoQ) in the dataset, and borrowings ticked up to $90.6 billion.Approximately 70% of CLF members also belong to the Federal Home Loan Bank system. The two facilities serve complementary purposes:
- FHLB: Market-rate wholesale funding backed by mortgage collateral, yielding ~6–7%. Used for ongoing balance sheet management and strategic leverage.
- CLF: Liquidity insurance of last resort backed by treasuries, yielding ~4%. Designed for the scenario where market sources of funding are "unavailable or impractical."
Chairman Hauptman emphasized this distinction: the CLF exists for tail risk. Its value is precisely in the scenario that CUs hope never materializes — an extended deposit drought or market dislocation where the FHLB or other counterparties cannot or will not lend. The Fed's September rate cut (25 bps to 4.00–4.25%) was announced the day after the board meeting; at that time, markets were pricing 125–150 bps of total cuts through December 2026.
Board Member Harper raised the legislative question of restoring the CLF's authority to lend through corporate credit unions to a subset of members — a capability that briefly existed under the CARES Act. Both Harper and Hauptman supported the concept as an additional layer of liquidity protection for smaller institutions.
Post-Exam Survey: A 41% Pain Point
The NCUA Ombudsman's post-examination survey achieved a 52% response rate in 2024 (up from 40% in 2023) and yielded 94% positive responses on exam report collaboration and delivery. But the sharpest finding was operational: 41% of credit union respondents reported that duplicate requests for data and information were not avoided during their exam. Credit unions noted that items uploaded to the secure portal were requested again by examiners on-site.
Examiners, for their part, indicated that the hybrid on-site/remote exam posture introduced communication challenges and slowed the process. Several examiner responses recommended returning to fully on-site exams, noting benefits for training new staff and improving communication with management.
The survey also surfaced examiner feedback that institutions under $100 million should receive more technical guidance from the NCUA — consistent with Chairman Hauptman's broader deregulation theme of reducing burden to help small credit unions survive and compete.
Participation patterns
Survey participation by credit unions has trended downward since the survey's inception in 2021, while examiner participation has risen (likely due to mandatory participation requirements). The Southern region had the highest participation rates in 2024; the Western region the lowest, likely attributable to fewer federal charters in that region. State-chartered credit unions contacted the Ombudsman's office after the survey launch to inquire about eligibility — suggesting potential demand for a broader survey scope.
Three Liquidations Since May
Chairman Hauptman acknowledged three credit union liquidations since the May board meeting, emphasizing that the timing was "coincidental" and "not a larger trend." The acting CFO identified common factors across 2025 failures: risk management weaknesses, operational deficiencies, persistent net losses, high operating expenses, and governance and leadership issues.
On a positive note, Hauptman highlighted several credit unions chartered in 2023–2024 that are surviving and growing, plus two in Minneapolis (Arise Community CU and Tribe FCU) that have provisional charters and are close to opening. New charters remain a priority for the agency — "it's not just to hand out a charter; it's important that these babies survive and learn how to walk."
FINASENSE Assessment
The July meeting's most consequential signal is the AI guidance gap. The GAO flagged it, the banking agencies' 2011 framework gives banks a clearer path, and the NCUA chose not to match it — opting for a resources page instead of formal guidance. This is a defensible regulatory choice (guidance becomes de facto requirements, and credit unions are diverse enough that one-size-fits-all AI rules could do more harm than good), but it leaves individual credit unions to navigate adoption decisions without a regulatory safe harbor. The institutions most affected are the Under $100M cohort — the ones least able to afford outside counsel to interpret the regulatory silence.
The CLF growth and the exam burden findings are operationally important but less strategically significant. The CLF signal is consistent with the tightening liquidity picture visible in the Q2 Call Report data. The 41% duplicate-request finding is a fixable process issue that the NCUA appears motivated to address.
