Domain Atlas / Lending & credit collections AI
Dave ExtraCash: an advertised ceiling, an automated amount, and a case that never asks how the amount is set
Explore this deployment in the PAN Lab ↗
In the PAN Lab, the readouts of this case's model organization carry a shaded evidence band whose width follows the least-established class among the modeling inputs the readings rest on.
The least-established input behind this case's model organization's readings is an assumption, not a measurement. Evidence base: 7 assumed · 5 published baseline.
In United States v. Dave, Inc. and Jason Wilk, No. 2:24-cv-09566-MRA-AGR (C.D. Cal.), the operative First Amended Complaint filed 30 December 2024 alleges that in the first fourteen months after Dave began advertising cash advances of 'up to $500', it offered a $500 advance to a new user about 0.002 per cent of the time — fewer than one determination in forty-five thousand; that only about 0.13 per cent of new users were offered even half of the advertised amount; that the most common offer, when an offer was made, was $25; that more than three-quarters of the time no advance was offered at all; and that on average more than 40 per cent of new users obtained no offer in a calendar month. Of new users who did receive offers, about 0.009 per cent of offers were for $500 and about 0.56 per cent were for at least $250; for existing users over the same window, on average more than a third were offered no advance in a calendar month and a $500 advance was offered less than 1 per cent of the time. THE DEFENDANTS DENY PARAGRAPHS 34, 35 AND 36 IN THEIR ENTIRETY and state in their dismissal brief that 'Dave can, does and did provide $500 advances'. The pleading never dates the fourteen-month window. Independently of the disputed tail, and undisputed because the operator publishes it, the advertised ceiling is $500 while Dave's own SEC-filed average advance was $170 in fiscal 2024 and $205 in fiscal 2025.[5]
What happened
Dave, Inc. is a publicly traded consumer financial services company in Los Angeles. Its cash-advance product, ExtraCash, is advertised with one number: up to $500, in five minutes or less. A member links their primary checking account, and an engine the company calls CashAI reads that account's transaction history and returns two things — whether the member is eligible, and how many dollars. A partner bank, Evolve Bank & Trust or Coastal Community Bank, originates the advance as a discretionary overdraft on a demand deposit account.
Everything known about that engine comes from the company itself. Its Form 10-K for fiscal 2025 states that it uses "our proprietary AI-powered underwriting system, CashAI" to "analyze a Member's checking account transaction data to determine eligibility and set the bank's credit approval amount", in a "fully automated process" that "requires no credit check and does not rely on FICO or credit bureau data", drawing on "hundreds of data points — including income patterns, spending behavior, and transaction history". The same filing says the model has "leveraged insights from over 180 million ExtraCash originations and billions of bank transactions" and that "the short average term of ExtraCash (approximately 11 days) creates rapid feedback loops, enabling iterative model refinement". In September 2025 the company announced CashAI v5.5, describing it as trained on more than 7 million recent originations that had reached full maturity, nearly doubling the prior feature set and optimized for the new fee structure. Its founder and chief executive, Jason Wilk, is quoted in that release calling the system "a powerful differentiator for Dave". Those are marketing and investor-relations claims. No third party has validated any of them.
The scale is the operator's own too: over 19 million members signed up since inception and over 14 million having used at least one product; more than $22 billion in ExtraCash provided; origination volume rising from approximately $5.1 billion in fiscal 2024 to approximately $7.6 billion in fiscal 2025; an average advance of $170 rising to $205; approximately 280 full-time employees at the end of 2025.
On 5 November 2024 the Federal Trade Commission sued. On 30 December 2024 it referred the matter to the Department of Justice, which filed the operative First Amended Complaint adding Wilk personally and a demand for civil penalties. The government's central quantitative allegation is a distribution claim rather than a fairness claim. It alleges that in the first fourteen months after Dave began advertising "up to $500", when determining whether and in what amount to offer an advance to a new user, Dave offered $500 about 0.002 per cent of the time — fewer than one determination in forty-five thousand; that only about 0.13 per cent of new users were offered even half of the advertised amount; that the most common offer, when an offer was made, was $25; that more than three-quarters of the time no advance was offered at all; and that on average more than 40 per cent of new users obtained no offer in a calendar month. Dave and Wilk deny paragraphs 34, 35 and 36 of the amended complaint in their entirety. Their dismissal brief states that "Dave can, does and did provide $500 advances" and that consumers "would not, as a matter of law, be misled to think that they were guaranteed to get $500".
Three of the five counts are not about the advance at all. They are Restore Online Shoppers' Confidence Act counts about a monthly membership fee charged to every consumer who linked a bank account, whether or not any advance was ever offered, and about the difficulty of stopping it: the pleading alleges at least nine separate in-app steps from the main screen to complete cancellation, diversion from cancellation for consumers who select the most prominent option, identity checks demanded to cancel including date of birth, sign-up phone number, mailing address, the last four digits of a Social Security number and details of the last two transactions on the external bank account, a July 2020 customer-service instruction that only consumers with no open advance and no pending advance payment were eligible to pause, and one consumer who required 27 days and nine messages to support (the court's order recites 29 days). A fourth count concerns the historic tip mechanic: a default charge of 15 per cent of the advance presented behind a large green "Thank you!" button above imagery of a cartoon child and boxes reading "10 Healthy Meals", "15 Healthy Meals", "20 Healthy Meals", with the alternative rendered white on white at about half the width, and the child replaced by an empty plate when the slider moved to zero. The pleading alleges Dave donated ten cents per percentage point of tip, usually $1.50 or less per advance, and kept the rest; Dave admits only that it "donated a portion of each tip" and denies the remainder. The FTC's November 2024 press release states, citing Dave's own Securities and Exchange Commission filings rather than the complaint, that Dave reported more than $149 million in revenue from tips from 2022 through the first six months of 2024.
On 12 September 2025 Judge Monica Ramirez Almadani denied the motion to dismiss in full in a 34-page order. On the "up to" theory the court held that "numerous courts have found that 'up to' representations can materially mislead reasonable consumers where the defendant does not or cannot provide the good or service as represented, especially when the representation references a particular, quantified amount", that "the government has plausibly alleged that it was exceedingly rare for Dave to offer the maximum amount of the cash advance advertised or even amounts approaching the maximum", and that a fine-print "Terms apply" disclaimer in two banner advertisements did not cure the net impression, because "a disclaimer does not automatically exonerate deceptive activities". That is a plausibility ruling on the pleadings and not a finding of fact — and the order's own footnote 2 is the reason this file treats the headline statistic as contested measurement: "Defendants contend that the government's method of calculating cash advances is wrong and that the data it used is [in]complete. ... Such a factual dispute cannot be resolved on a Rule 12(b)(6) motion to dismiss."
Dave answered on 10 October 2025 with a general denial and seven affirmative defences, including lack of fair notice of what it calls the government's novel interpretation of the statute, a vagueness challenge to that statute as applied, mootness, good faith resting in part on an assertion that "the Consumer Financial Protection Bureau opened and closed an investigation — and declined to recommend an enforcement action against Dave" (an assertion by the defendant; no agency document confirming it was located), offsets, limitations, and a challenge to the penalties sought as unconstitutionally excessive. Publicly, the company called the amended complaint "a continued example of government overreach" resting on "numerous allegations that are based on various inaccuracies", said it believes it has "always acted within the law", and pledged to "vigorously defend itself".
Two things moved while the case ran. The price surfaces were replaced: members onboarded from 4 December 2024 were put on a structure without optional tips or express fees, and in February 2025 the company completed a transition to a mandatory 5 per cent overdraft service fee with a $5 minimum, with tip revenue falling 89 per cent from $67.6 million to $7.5 million year over year. And the disclosure moved. The footnote the amended complaint quotes from Dave's site shortly after the November 2024 filing said only that "the average advance is $170" and that enrolment and initial qualification are typically completed in five minutes. The same site as displayed on 28 August 2026 still leads with "Up to $500 in 5 min or less", and its footnote now reads: "ExtraCash amounts range from $25-$500, typically authorized within 5 minutes, with an overdraft fee equal to the greater of $5 or 5%. Multiple overdrafts may be required. Not all members qualify for ExtraCash and few qualify for $500." The two concessions now present in the fine print — that not all members qualify at all, and that few qualify for the maximum — are precisely the two omissions the amended complaint pleads.
The 2025 change of FTC leadership did not end or shrink the matter. The Department of Justice, real party in interest since December 2024, filed the opposition to dismissal on 7 April 2025 under the new administration, argued it on 30 June 2025 and won it; the Commission's own case page for matter 232 3014 still records "Case Status: Pending". Two private actions run alongside: a putative class action under the Military Lending Act and the Truth in Lending Act naming Dave and Evolve Bank & Trust, in which dismissal and arbitration were both denied on 12 December 2025 and which is now before the Ninth Circuit with district proceedings stayed; and a suit by the Mayor and City Council of Baltimore under a municipal consumer-protection ordinance, removed to federal court in January 2026. Dave's Form 10-Q filed 5 August 2026 says the company is "unable to reasonably predict the possible outcome" and records a $9.7 million aggregate accrual for legal contingencies across its three pending consumer matters. Nothing is settled and nothing is decided.
The sociotechnical reading
The governed surface here is the gap between one advertised number and the distribution an automated system actually emits — and the striking feature of the record is that the enforcement half and the system half never touch.
Start with the absence, because it is the finding. The words algorithm, artificial intelligence, machine learning, model and underwriting appear nowhere in the original complaint, the operative amended complaint, the defendants' dismissal memorandum, the government's opposition, the answer, or the court's 34-page order. The pleading describes only that Dave "uses its access to consumers' bank accounts to analyze their finances and banking history" and "uses this information to make decisions about how much (if any) to advance the consumer". The government's theory is a marketing theory. The decision system is a black box whose outputs are counted, and the party that named it, versioned it and published its training scale is the operator, in its own investor filings. So this is not a case in which a regulator challenged an automated decision system; it is a case in which one was counted from the outside while nobody asked what was inside. There is no model documentation, no validation report, no fairness assessment, no accuracy audit and no independent evaluation of this engine anywhere in the public record.
Second, the enrolment order is the mechanism, and it runs backwards from the way a lending decision is usually drawn. Everything the member surrenders is surrendered BEFORE the engine answers: read access to their primary checking account at the tap that links it, a recurring monthly subscription, and historically an express fee to receive the money immediately and a default charge presented as a tip. The pleading alleges that on average more than 40 per cent of new users obtained no offer at all in a calendar month and were charged the monthly fee regardless. On that account the funnel monetises the arrival rather than the approval, which is why three of the five counts are about a subscription and a cancellation path rather than about an amount.
Third, the disputed object is a statistic about the system rather than the system's accuracy, and both sides are content to leave the accuracy question alone. Nobody argues the engine is wrong. The operator advertises it as improving and reports approval rates at all-time highs and higher average approval amounts after v5.5; the government counts outputs and never asks how they are produced; the court records that the parties cannot agree on how the counting may lawfully be done. That makes this a different governance object from a fairness challenge to a screening model, and a different one again from a model whose accuracy is itself the harm. Here the model is treated as sound by both sides, and the contested thing is a description of what it returns. One reading of the gap needs no telemetry at all and nobody disputes it: an advertised ceiling of $500 against the operator's own published average advance of $170 in fiscal 2024 and $205 in fiscal 2025.
Fourth, the feedback asymmetry. On the operator's side, repayment outcomes return on an eleven-day cycle and are used for retraining; interface variants are measured against conversion and revenue; the pleading alleges an experiment removing the charitable-meal content reduced the share of new users charged a tip by about a third and cut tip revenue by almost a quarter, that an internal analysis recommended resuming the content for all users, and that a second experiment removing another screen likewise reduced both the number charged and the amounts. That is a measurement apparatus that can price an interface change within days. On the other side, the consumer-protection loop ran on a two-year investigation clock and is, four years from the advertising conduct, at a pretrial conference with no ruling on the merits. Dissatisfaction was measured on the operator's side with real precision — internal analyses naming "Low advance amount", "Low advance limits and approval" and "Advance request denied" among the top drivers of contact, a survey naming "Not enough money" a top source of dissatisfaction, a presentation stating that on the express-fee screen "what we promised is not what they see" — and the pleading alleges the recommendations those analyses carried each went unimplemented. All of that is denied, and Dave refers the court to the documents in their entirety.
Fifth, discretion sits entirely upstream and the decision subject has none of it. The consumer cannot appeal an amount, cannot see why it was set, and is given no adverse-action explanation; the operator's own description is a fully automated process. The only move available is to reopen the app and be scored again. Support staff can neither raise an amount nor, on the pleaded account, reliably stop a recurring charge. Everything that could change is held by the team that sets the model, the ceiling and the interface — which is also the team the internal analyses reached.
Finally, the one documented repair, and it is worth naming precisely because it is small and real. The disclosure moved during the litigation, and the two concessions it gained are exactly the two omissions the pleading alleges. The instrument that produced that change was not a model control or an audit; it was a published page being read closely by someone outside the company, four years and one federal complaint later. Reading this deployment honestly means holding both halves at once: an engine nobody outside has ever examined, whose accuracy nobody disputes, and a consumer-protection record that has moved a fine print and nothing else.
The concepts used in this reading are defined in the Field Guide; the governance responses live in the Practice Library. The model organization for this case can be stress-tested in the PAN Lab.