Every earned wage access app says the same three words on its homepage: not a loan. For years that was a marketing position, not a legal one, because nobody, including the products making the claim, could say for certain whether federal law agreed. On December 23, 2025 the CFPB ended the ambiguity with an advisory opinion that draws an exact, four-part line around which products get to make that claim honestly. Almost none of the popular consumer apps clear it, and the reason they don't is not an accident of paperwork. It is a direct result of how they are built.
Here is the plain version. Earned wage access lets a worker draw down money they have already earned before the scheduled payday arrives. That is the pitch everywhere. What decides whether it is credit under the Truth in Lending Act is not the pitch, it is the mechanism underneath it: where the data comes from, how repayment happens, and what happens when repayment falls short.
The CFPB drew the line in December
The advisory opinion, published in the Federal Register as 90 FR 60069, defines a narrow category it calls Covered EWA and states plainly that Covered EWA is not credit under Regulation Z. To earn that status a product has to clear four specific conditions at once.
The transaction cannot exceed the wages the worker has already earned, and that amount has to come from payroll data, not from the worker's own estimate or a lender's prediction of what they are likely to have earned. Repayment has to happen through an actual payroll-process deduction tied to the worker's next payroll event, not a debit pulled from whatever bank account the worker connected. The provider has to warrant, as part of the contract, that it has no recourse if that payroll deduction falls short: no collections activity, no selling the shortfall to a third party, no reporting it to a credit bureau. And the provider cannot assess an individual worker's credit risk at all, not through a credit pull, not through a proprietary score, nothing that looks like underwriting a person rather than reading a payroll record.
Clear all four and fees stop being a Regulation Z question at all, since a finance charge is the cost of credit and Covered EWA is not credit. The more useful half of the opinion covers everything else. For EWA products that do count as credit, optional expedited-delivery fees and voluntary tips still are not finance charges in the normal course, because the consumer opts into them rather than having them imposed. That holds only while the choice is real: make the standard free delivery path too hard to pick, the opinion says, and those fees may effectively be imposed, which turns them back into finance charges. Missing one of the four conditions means the product gets evaluated as ordinary consumer credit, but it does not by itself take the fee model down with it.
Almost nothing on your phone is built this way
Look at what actually sits behind the "not a loan" homepage on the products people use every day, and the pattern is consistent. The advance limit comes from a proprietary cash-flow model reading bank-linked transaction history, not a payroll feed. Underwriting is real, it is just alternative data instead of a credit bureau: a behavioral score, a bank-account age and deposit-consistency check, a recurring-direct-deposit signal, something built to size a limit and decide who qualifies. Repayment is an ACH pull from the connected bank account on a date the product predicts, not a payroll-system deduction the employer executes.
That combination, bank data plus an individual risk model plus consumer-account repayment, is precisely the shape the fourth prong excludes. It does not make these products illegal. It makes them ordinary consumer financial products that have chosen, deliberately, to compete on speed and low friction rather than on qualifying for the narrow federal carve-out, and they are underwritten and priced accordingly, usually through membership fees, optional express-delivery charges, and voluntary tips rather than interest.
The employer-partnered model is the other path
A smaller set of products take the harder route: direct integration with an employer's payroll system, reading the same payroll data the employer already runs, and repaying through the same payroll process rather than a separate consumer debit. That path costs more to build, because it means selling into and integrating with an employer rather than acquiring a consumer directly, but it is the one the December opinion was actually written to protect. No individual risk assessment, no consumer-account recourse, and the accrued-wage basis is a fact from the payroll system rather than an estimate.
The choice between the two is not really a legal question dressed up as an engineering one. It is a genuine product decision about who your customer is, and it decides almost everything that comes after it.
What actually has to be built, on either side
If the target is Covered EWA, the requirement is architectural, not cosmetic. The system needs a real integration into payroll data, either an API or a file-based feed a payroll provider or employer supplies, so that the accrued-wage figure is a fact you can point to rather than a number your model produced. Repayment has to route through an actual payroll-process deduction, which means the employer relationship is part of the product, not an afterthought. And the no-recourse warranty has to be a real contractual term the product operates by, not a line in the terms of service that the collections flow quietly ignores. Add a risk score anywhere in that pipeline, even a soft one meant to catch edge cases, and the product no longer qualifies, because the fourth prong is unconditional.
If the target is the bank-data model instead, the honest engineering center of gravity moves elsewhere. Cash-flow underwriting is the real product, not a footnote, and pretending otherwise is how teams end up with a limit-sizing model nobody can defend when a regulator or a disgruntled user asks how a number was set. The repayment side needs the same idempotent, event-sourced discipline as any other money movement, so a retried debit never double-charges and every attempt has an audit trail. And the fee design has to make optionality real rather than theoretical: a genuinely usable free path, defaults that do not quietly steer toward the paid one, and a fee schedule shown before the user commits, because the December opinion's finance-charge carve-out for expedited fees and tips depends on exactly that being true in practice, not just in the copy.
We build the underwriting, repayment, and disbursement machinery behind consumer lending products, the cash-flow models, the event-sourced ledgers, and the audit trail that holds up when someone asks how a limit or a decision was reached, which is the same event and correctness discipline every money-moving product needs regardless of which side of this line it sits on. The behavioral risk-scoring engine that many of these apps use to size limits and detect risk without a bureau pull is its own build, with its own version of the same event-sourced discipline.
What 2muchcoffee covers
We build the systems behind lending and cash-advance products, whichever side of the Covered EWA line a founder is building toward: the payroll integration and no-recourse contract logic for a true EP model, or the cash-flow underwriting and idempotent repayment engine for a bank-data product, plus the audit trail that makes either one defensible later. If your product says "not a loan" and you are not certain yet which of the four conditions it would fail, that is the conversation worth having before launch. The plain way in is the AI and engineering work we do.
FAQ
What is earned wage access? A product that lets a worker draw down wages they have already earned before their scheduled payday, rather than borrowing against future income. Whether a specific product counts as a loan under federal law depends on how it is built, not on how it is marketed.
What is Covered EWA under the CFPB's December 2025 opinion? A narrow category of earned wage access that the CFPB says is not credit under Regulation Z. To qualify, a product cannot exceed wages evidenced by payroll data, must repay through an actual payroll-process deduction, must warrant no recourse if that deduction falls short, and must not assess individual credit risk at all.
Why do most earned wage access apps not qualify as Covered EWA? Because they use bank-linked cash-flow data and a proprietary risk model to set limits, rather than payroll data, and they repay through a consumer bank-account debit rather than a payroll-process deduction. That combination fails the fourth prong of the CFPB's test, which requires no individual credit-risk assessment.
Are expedited fees and tips finance charges under the new opinion? Not if the free or standard path is genuinely usable and the paid path is not the only practical option. The CFPB's opinion ties that conclusion to the fee being truly optional in practice, not just described as optional in the terms.
What is the employer-partnered EWA model? A model that integrates directly with an employer's payroll system, reading real payroll data and repaying through the same payroll process, rather than acquiring a consumer directly and debiting a connected bank account. It costs more to build but is the model the December 2025 opinion was written to protect.
Does "no credit check" mean an earned wage access app has no underwriting? No. Every consumer-facing app in this space runs a real risk model, it just uses bank-account signals, such as deposit consistency and account age, instead of a credit bureau pull. The absence of a credit check is not the absence of underwriting.
One concrete action
Take your own product's "not a loan" claim and check it against the four conditions one at a time: accrued wages from payroll data, repayment via payroll-process deduction, a real contractual no-recourse warranty, and zero individual risk assessment. The moment any one of them fails, you are not building Covered EWA, you are building consumer credit with a friendlier front end, and every downstream decision, from disclosures to how you price the fee, changes because of it.
