BRENNEN

Industries / Finance and business services / Finance and banking / Credit unions

Lending

Credit unions lend members' savings back to members, mostly for cars, homes and personal needs, and some lend to members' businesses.

5

Milestones achieved, from 1998 to 2022.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1998

    A cap on business loans

    NCUA first regulated member business loans in 1987. The Credit Union Membership Access Act of 1998 then capped a credit union's total member business loans at the lesser of 1.75 times its net worth (its retained capital) or 1.75 times the net worth needed to count as well capitalized. Credit unions with a history of mainly making business loans before 30 September 1998 kept an exception from the cap.

    govinfo.govncua.gov

    Credit

    the US Congress

  2. 2010

    A payday loan alternative

    The Federal Credit Union Act caps the annual percentage rate on federal credit union loans at 15%, all finance charges included, unless NCUA raises it. NCUA's 2010 rule let federal credit unions make payday alternative loans (PALs) of $200 to $1,000, repaid in full in installments over one to six months with no rollovers, at an annual rate up to 10 percentage points above NCUA's general ceiling, 28% at the time, plus an application fee of at most $20. In September 2011, 372 federal credit unions offered PALs, with $13.6 million outstanding on 36,768 loans; by December 2018, 606 offered them, with $145.2 million on 211,589 loans. A 2019 rule added PALs II, of up to $2,000 for one to twelve months, open to a member from the day they join.

    govinfo.govgovinfo.gov

    Credit

    NCUA Board

  3. 2016

    Business loans by principle

    NCUA replaced its detailed rules on member business loans, such as fixed collateral and personal guarantee requirements, with broad principles, mostly from 1 January 2017, and ended the waivers credit unions had needed to depart from those rules. It had more than 1,000 such waivers in force and processed 336 in 2014 and 225 in 2015. The statutory cap stayed.

    govinfo.gov

    Credit

    NCUA Board

  4. 2018

    Taxi medallion loans fail

    A taxi medallion is a transferable city permit to run a taxi. New York City medallions sold for more than $1 million in 2013, and ride-hailing companies such as Uber and Lyft entered from 2012. Three New York credit unions that lent heavily against medallions, Melrose, LOMTO and Bay Ridge, were taken over by regulators or merged away in 2017 and 2018, at an estimated cost of $765.5 million to the National Credit Union Share Insurance Fund. At Melrose, which held the exception from the business loan cap, medallion loans were about 71% of its loans in June 2018. NCUA's Office of Inspector General found that the three failed from concentration in medallion loans, unsafe lending and weak oversight by boards and managers.

    Lessonan exemption from a concentration limit let one kind of collateral take over a lender's book.

    ncua.gov

    Credit

    NCUA Office of Inspector General, with the auditors Moss Adams LLP

  5. 2022

    Credit builder loans tested

    In a credit builder loan the lender puts the loan amount in a locked savings account and releases it as the borrower makes payments, which are reported to credit bureaus like any loan. At St. Louis Community Credit Union, a certified community development financial institution (CDFI) that had offered such loans since 2009, the loan was $600 repaid in 12 monthly payments of about $54. Researchers randomly varied how easy it was for 1,531 interested members to take one in 2014–2015. Credit scores did not change on average: people with no loan on their credit report tended to gain, while people who already had loans tended to fall behind on those loans and lose points.

    Lessona loan meant to build credit can hurt borrowers who are already stretched, so lenders need to screen who takes it.

    rand.orgpovertyactionlab.org

    Credit

    Jeremy Burke, Julian Jamison, Dean Karlan, Kata Mihaly and Jonathan Zinman, in The Review of Financial Studies, 2022

Watch for

  1. AI underwriting for small lenders

    Commonwealth Credit Union and Zest AI launched CU Lending Collective, a credit union service organization (a company owned by credit unions to serve them) that offers small credit unions AI-based loan underwriting with fair-lending documentation and reason codes. Commonwealth says its loan growth exceeded 14% in 2025 after it began using Zest AI, with strong delinquency and charge-off performance, and Zest AI's chief executive says the technology is used by hundreds of credit unions.

    Achieved whenan outside study on public call report data shows that credit unions using such underwriting approve more members or grow loans faster than similar credit unions without it, with no rise in delinquency or charge-offs.

    autoremarketing.comccuky.org

    Claimed by

    Commonwealth Credit Union and Zest AI, February 2026

Open

  1. 01

    AI lending claims, checked

    An outside test of whether credit unions that adopted AI loan underwriting changed their lending results. Take credit unions whose adoption date is public, from vendor case studies and press releases, and compare their auto and personal loan growth, delinquency (loans 30 or more days late) and net charge-offs (loans written off, less recoveries) quarter by quarter before and after adoption, against similar credit unions that announced no adoption.

    The brief
    Credit · unclaimed
  2. 02

    Failure flags ranked by cost

    A test of public early-warning measures against the cost of credit union failures, not only their number. From call report data, compute simple measures such as the share of loans in the largest single category, business loans against net worth and loan growth, plus a standard failure model, and score what share of the share insurance fund's losses since 2010 each would have flagged four to eight quarters ahead. Publish results in aggregate, without naming open credit unions.

    The brief
    Credit · unclaimed

Open 01 · the brief

AI lending claims, checked

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An outside test of whether credit unions that adopted AI loan underwriting changed their lending results. Take credit unions whose adoption date is public, from vendor case studies and press releases, and compare their auto and personal loan growth, delinquency (loans 30 or more days late) and net charge-offs (loans written off, less recoveries) quarter by quarter before and after adoption, against similar credit unions that announced no adoption.

Why it is still open

The evidence so far is the vendors' and their clients' own figures. A November 2026 review of call report data for 4,203 credit unions found "no significant credit-quality improvement" from AI so far, but its authors had no systematic measure of which credit unions had adopted AI and wrote that the comparison cannot establish causation. Golden 1's chief executive reported a 28% rise in approvals for protected classes of borrowers without describing how it was measured. A search found no study that matches named adopters to peers.

cuinsight.combankingdive.com

Built with what exists

NCUA's quarterly call report files, which give each credit union's loan balances, delinquencies and charge-offs by loan type; adoption dates from Zest AI's and other vendors' published case studies; Python with pandas, and the R package did for staggered difference-in-differences.

ncua.gov

  1. First step

    List the credit unions named in Zest AI's public case studies, date each adoption from its announcement, download the 2019–2026 call report files and plot the series for the first five.

  2. Done looks like

    At least 20 adopting credit unions with a dated adoption, each matched to 5 peers by asset size, state and loan mix; the estimated change in 30-day delinquency and net charge-off rates on auto and personal loans 4 and 8 quarters after adoption, with 95% confidence intervals; code and the dated adopter list published.

Open 02 · the brief

Failure flags ranked by cost

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A test of public early-warning measures against the cost of credit union failures, not only their number. From call report data, compute simple measures such as the share of loans in the largest single category, business loans against net worth and loan growth, plus a standard failure model, and score what share of the share insurance fund's losses since 2010 each would have flagged four to eight quarters ahead. Publish results in aggregate, without naming open credit unions.

Why it is still open

A 2025 study by Qiao Peng, Donal McKillop, Barry Quinn and Kailong Liu predicts US credit union liquidations a year ahead with 97.9% test accuracy using 44 factors, a measure that counts each failure the same, small or large. The costliest recent failures were three taxi medallion lenders whose loss was estimated at $765.5 million and whose cause was concentration. A search found no public test that weights early warnings by the losses they would have caught.

pure.qub.ac.ukncua.gov

Built with what exists

NCUA's quarterly call report files; the losses estimated in NCUA's Inspector General material loss and limited reviews; Python with scikit-learn.

ncua.govncua.gov

  1. First step

    Tabulate the loss estimates in the Inspector General's reviews since 2010 and download the call report files for 2012 to 2018.

  2. Done looks like

    A list of every federally insured credit union failure from 2010 to 2025 with its estimated loss; for each measure, the share of total losses flagged at least four quarters before failure while flagging no more than 5% of open credit unions each quarter; code and aggregate results published.