# Massachusetts fined Earnest $2.5M over AI underwriting that penalised Black, Hispanic and non-citizen student-loan applicants (2025)

> On 10 July 2025 Massachusetts Attorney General Andrea Joy Campbell announced a $2.5 million settlement with student-loan lender Earnest Operations LLC, resolving allegations that its AI/algorithmic underwriting models could disparately harm Black, Hispanic and non-citizen applicants: a 'Cohort Default Rate' variable that penalised Black and Hispanic applicants more than White ones, and a 'Knockout Rule' that auto-denied anyone without at least a green card. Every figure is quoted verbatim from the AG's own release and independent trade press.

- Verification status: pending
- Case type: deployment
- Provider: Earnest Operations LLC (the lender's own AI/algorithmic underwriting models)
- Client: Earnest Operations LLC — the Delaware-based student-loan lender that was the subject of the enforcement action, fintech (named)
- Sector: fintech / US / finance
- Canonical URL: https://theinternetninja.com/stories/earnest-operations-mass-ag-ai-underwriting-fair-lending-2-5m-settlement-2025/
- Source: The Internet Ninja (theinternetninja.com), independent verified-proof platform

## Outcomes

| Metric | Before | After |
| --- | --- | --- |
| Massachusetts AG settlement with Earnest Operations LLC (10 July 2025) | Algorithmic underwriting models the AGO alleged were not tested for disparate impact | $2.5 million paid to the Commonwealth, plus mandated AI-governance changes (Mass. AGO, 2025) |
| 'Cohort Default Rate' variable in the algorithmic model | A weighted subscore tied to the average default rate at an applicant's college | Disparate impact in approval rates and loan terms; Black and Hispanic applicants more likely penalised than White; variable discontinued (Mass. AGO, 2025) |
| Immigration-status 'Knockout Rule' | Automatic denial of applicants without at least a green card | Rule discontinued under the settlement (Mass. AGO, 2025) |

## Verification method

Every load-bearing figure is quoted verbatim from a source fetched, grepped and saved to sources/ this session. The $2.5M amount, the 10 July 2025 date, the disparate-harm allegation, the 'Cohort Default Rate' variable and its Black/Hispanic disparate impact, and the immigration-status 'Knockout Rule' are quoted from the Massachusetts Attorney General's Office press release (Tier 1, government primary; retrieved via the Wayback Machine capture 20260611170703 and saved to sources/) and independently corroborated by ABA Banking Journal (Tier 2, independent trade press; Wayback 20260307145847, saved to sources/). The settlement instrument — the assurance of discontinuance filed in Suffolk County Superior Court — is archived to sources/ as the underlying primary document. This is a public state-enforcement action; no confirmation was sought from Earnest and none is needed.

## FAQ

**What did Massachusetts allege about Earnest's AI underwriting?**

The Massachusetts Attorney General's Office alleged that Earnest Operations LLC used artificial-intelligence/algorithmic underwriting models to make lending decisions and failed to take reasonable measures to mitigate fair-lending risks, including failing to test the models for disparate impact. The office said the practices could lead to disparate harm to Black, Hispanic and non-citizen applicants and borrowers.

**How much did Earnest pay, and when was the settlement announced?**

Earnest Operations LLC agreed to pay $2.5 million to the Commonwealth of Massachusetts. Attorney General Andrea Joy Campbell announced the settlement on 10 July 2025, reached via an assurance of discontinuance filed in Suffolk County Superior Court.

**What were the 'Cohort Default Rate' variable and the 'Knockout Rule'?**

The Cohort Default Rate was a variable in Earnest's algorithmic model reflecting the average loan-default rate at an applicant's educational institution; the AGO said its use resulted in disparate impact in approval rates and loan terms, with Black and Hispanic applicants more likely to be penalised than White applicants. The Knockout Rule automatically denied applications based on immigration status, denying applicants who lacked at least a green card. Under the settlement Earnest discontinues both.

## Full case file

**Verification status: CHECKING — the maker has quoted every critical figure verbatim from the Massachusetts AG's own release and corroborated it against independent trade press; awaiting the checker's audit before any badge.**

## The problem
Earnest Operations LLC is a Delaware-based student-loan lender, and the system at the centre of this case is not a product it sold but the machinery it used to decide who got a loan ([source](https://www.mass.gov/news/ag-campbell-announces-25-million-settlement-with-student-loan-lender-for-unlawful-practices-through-ai-use-other-consumer-protection-violations)). According to the Massachusetts Attorney General's Office, "Earnest uses artificial intelligence models, specifically algorithmic models, to make lending decisions, including determinations on loan applicants' eligibility and loan terms and pricing" ([source](https://www.mass.gov/news/ag-campbell-announces-25-million-settlement-with-student-loan-lender-for-unlawful-practices-through-ai-use-other-consumer-protection-violations)). The question a fair-lending enforcer asks of such a system is whether it was tested for who it disadvantages, and the AGO alleged it was not: that Earnest "failed to take reasonable measures to mitigate fair lending risks in its underwriting practices, including failing to test its models for disparate impact and training its models based on arbitrary, discretionary human decisions" ([source](https://www.mass.gov/news/ag-campbell-announces-25-million-settlement-with-student-loan-lender-for-unlawful-practices-through-ai-use-other-consumer-protection-violations)).

## What was built
What Earnest built was an automated underwriting pipeline, and independent trade reporting dates and describes it: ABA Banking Journal reports that "Since 2014, Earnest used AI-based underwriting models to issue personal and student loans through a three-stage algorithmic process that applied knockout rules" ([source](https://bankingjournal.aba.com/2025/08/mass-ag-reaches-settlement-with-earnest-operations-for-2-5m-over-ai-lending-bias/)). Two features of that pipeline became the load-bearing facts of the case. The first was a scoring input tied to schools: the AGO described "the company's use of the 'Cohort Default Rate' – an average rate of loan defaults associated with a specific educational institution – variable in its algorithmic model," which ABA Banking Journal renders as "a weighted subscore based on the CDR, which reflected the average loan default rate at an applicant's college" ([source](https://bankingjournal.aba.com/2025/08/mass-ag-reaches-settlement-with-earnest-operations-for-2-5m-over-ai-lending-bias/)). The second was a hard filter on immigration status, which the AGO called a "'Knockout Rule' to automatically deny applications based on immigration status" ([source](https://www.mass.gov/news/ag-campbell-announces-25-million-settlement-with-student-loan-lender-for-unlawful-practices-through-ai-use-other-consumer-protection-violations)).

## The outcome
The outcome was a state-enforcement settlement with a specific price. On 10 July 2025 the Massachusetts Attorney General's Office announced that "Attorney General Andrea Joy Campbell today announced that her office has reached a <span class="kpi">$2.5 million</span> settlement with Earnest Operations LLC (Earnest), a Delaware-based student loan company, resolving allegations that the company's lending practices violated various consumer protection and fair lending laws, including through the use of artificial intelligence (AI) models that could lead to disparate harm to Black, Hispanic, and non-citizen applicants and borrowers" ([source](https://www.mass.gov/news/ag-campbell-announces-25-million-settlement-with-student-loan-lender-for-unlawful-practices-through-ai-use-other-consumer-protection-violations)). Independent trade press states the same figure and effect: "Student loan company Earnest Operations LLC agreed to pay <span class="kpi">$2.5 million</span> to the Commonwealth of Massachusetts to resolve allegations it used artificial intelligence models that disproportionately harmed Black and Hispanic applicants" ([source](https://bankingjournal.aba.com/2025/08/mass-ag-reaches-settlement-with-earnest-operations-for-2-5m-over-ai-lending-bias/)).

What the model actually did to applicants is the part that matters, and the AGO stated it plainly. On the schools variable, the office said its use "resulted in disparate impact in approval rates and loan terms for a certain product, with Black and Hispanic applicants more likely to be penalized than White applicants" ([source](https://www.mass.gov/news/ag-campbell-announces-25-million-settlement-with-student-loan-lender-for-unlawful-practices-through-ai-use-other-consumer-protection-violations)). On the immigration filter, ABA Banking Journal reports the AGO alleged Earnest "using a knockout rule to automatically deny applicants who lacked at least a green card, creating an ECOA disparate impact risk" ([source](https://bankingjournal.aba.com/2025/08/mass-ag-reaches-settlement-with-earnest-operations-for-2-5m-over-ai-lending-bias/)). Attorney General Campbell tied the harm to the borrowers it fell on: "Earnest's failure to comply with consumer protection and fair lending laws, including through its AI models, unfairly put historically marginalized student borrowers at risk of being denied loans or receiving unfavorable loan terms – impeding their chances of economic growth and opportunity" ([source](https://www.mass.gov/news/ag-campbell-announces-25-million-settlement-with-student-loan-lender-for-unlawful-practices-through-ai-use-other-consumer-protection-violations)).

The settlement did more than levy a payment; it changed the model. It "was reached via an assurance of discontinuance filed in Suffolk County Superior Court," and under it "Earnest will implement a detailed corporate governance structure and develop and maintain robust written policies to ensure responsible and legally compliant use of AI," and "will also discontinue use of both the 'Cohort Default Rate' variable and 'Knockout Rule' based on immigration status" ([source](https://www.mass.gov/news/ag-campbell-announces-25-million-settlement-with-student-loan-lender-for-unlawful-practices-through-ai-use-other-consumer-protection-violations)). One boundary is worth stating: these are the Attorney General's allegations, resolved by settlement rather than by a court's merits finding of discrimination, and the assurance of discontinuance is the instrument that closed them.

**Weakest load-bearing source.** The strongest source here is also the accuser: the $2.5 million figure and the descriptions of the Cohort Default Rate and Knockout Rule come from the Massachusetts Attorney General's Office itself, a government primary that is a party to the action rather than a neutral third party ([source](https://www.mass.gov/news/ag-campbell-announces-25-million-settlement-with-student-loan-lender-for-unlawful-practices-through-ai-use-other-consumer-protection-violations)). That is why this page leans on independent corroboration for the same facts from ABA Banking Journal, which is not a party to the case ([source](https://bankingjournal.aba.com/2025/08/mass-ag-reaches-settlement-with-earnest-operations-for-2-5m-over-ai-lending-bias/)). The underlying settlement instrument, the assurance of discontinuance filed in Suffolk County Superior Court, is archived to sources/ as the primary document, but its full text was not machine-read this session, so every quotation on this page is taken from the AGO release or the trade report, never from unread PDF numerals. No confirmation was sought from Earnest; the record is already public.

> **How this was verified.** Method: every load-bearing figure was retrieved live this session, grepped for the exact string, and saved to sources/. The $2.5 million amount, the 10 July 2025 date, the "disparate harm to Black, Hispanic, and non-citizen applicants" framing, the "Cohort Default Rate" variable and its "Black and Hispanic applicants more likely to be penalized than White applicants" disparate impact, the immigration-status "Knockout Rule," AG Campbell's quote, and the assurance-of-discontinuance settlement terms are quoted verbatim from the Massachusetts Attorney General's Office press release (Tier 1, government primary), read through the Wayback Machine capture 20260611170703 and saved to sources/massgov-ag-earnest-release.html. Each is independently corroborated by ABA Banking Journal (Tier 2, independent trade press), read through Wayback capture 20260307145847 and saved to sources/aba-bankingjournal-earnest.html, which also supplies the "Since 2014… three-stage algorithmic process" and "green card" detail. The settlement instrument itself — the assurance of discontinuance filed in Suffolk County Superior Court — is archived to sources/earnest-assurance-of-discontinuance-suffolk-superior.pdf as the underlying primary document (text not machine-read this session; no numerals quoted from it). No confirmation was sought from Earnest. Checked 2026-09-03 (checking round 1).

## Related case files
- [SafeRent settled a class action over its algorithmic tenant-screening scores for $2.275M](/stories/louis-v-saferent-algorithmic-tenant-screening-2-275m-class-settlement-dmass-2024/) — the closest sibling from the same state and courthouse cluster: an automated scoring system challenged for discriminatory impact and resolved with a payment plus changes to the model, rather than by the vendor's own account.
- [Mobley v. Workday: a court cleared an AI hiring tool to be challenged at collective-action scale](/stories/mobley-v-workday-nd-cal-ai-hiring-tools-age-discrimination-nationwide-adea-collective-certified-2025/) — the same accountability shape one decision earlier in its life: an algorithmic decision system accused of disparate impact, made visible through a public legal process instead of marketing.
- [EEOC's $365K consent decree over iTutorGroup's AI hiring screen that auto-rejected older applicants](/stories/eeoc-itutorgroup-ai-hiring-age-screening-365k-consent-decree-2023/) — another automated "knockout" rule, this time on age, resolved by an enforcer and priced, showing the pattern is not confined to lending.

## Sources
1. Massachusetts Attorney General's Office · *AG Campbell Announces $2.5 Million Settlement With Student Loan Lender For Unlawful Practices Through AI Use, Other Consumer Protection Violations* · 10 July 2025 · https://www.mass.gov/news/ag-campbell-announces-25-million-settlement-with-student-loan-lender-for-unlawful-practices-through-ai-use-other-consumer-protection-violations — **Tier 1** (government primary; the enforcing party. The $2.5M amount, the 10 July 2025 date, the disparate-harm framing, the "Cohort Default Rate" variable and its Black/Hispanic disparate impact, the immigration-status "Knockout Rule," AG Campbell's quote and the assurance-of-discontinuance terms; saved to sources/massgov-ag-earnest-release.html, Wayback 20260611170703).
2. Earnest Operations LLC · *Assurance of Discontinuance* · filed in Suffolk County Superior Court, 2025 · https://www.mass.gov/doc/earnest-aod/download — **Tier 1** (primary settlement instrument; archived to sources/earnest-assurance-of-discontinuance-suffolk-superior.pdf. Text not machine-read this session; listed as the underlying document, no figures quoted from it here).
3. ABA Banking Journal · *Mass. AG reaches settlement with student loan firm for $2.5M over AI lending bias* · 1 August 2025 · https://bankingjournal.aba.com/2025/08/mass-ag-reaches-settlement-with-earnest-operations-for-2-5m-over-ai-lending-bias/ — **Tier 2** (independent trade press, not a party; corroborates the $2.5M figure, the Cohort Default Rate subscore, the green-card knockout rule and the "Since 2014… three-stage algorithmic process" detail; saved to sources/aba-bankingjournal-earnest.html, Wayback 20260307145847).