# Regulation Z 1026.36 Loan Originator Compensation and Where AI Vendor Pricing Sits: The Terms-Based Comp Prohibition, the Person-Definition Question, and How the Vendor Contract Has to Read

*July 10, 2026 · 12 min read · Ramkumar Venkataraman*

> The Loan Originator Compensation rule at Reg Z 1026.36 prohibits paying an individual loan originator based on the terms of the loan, and it defines 'loan originator' broadly enough to swallow the AI vendor conversation. The person-definition analysis, the compensation-attribution question that a per-loan vendor fee raises, and the contract structure that keeps the AI vendor's economics outside the LO Comp perimeter.

## The Rule That Reads Broader Than the Vendor Contract Expected

The Loan Originator Compensation rule at [12 CFR 1026.36(d)](https://www.consumerfinance.gov/rules-policy/regulations/1026/36/) is the CFPB's implementation of the Dodd-Frank prohibitions on steering incentives in mortgage origination. The rule prohibits any person from paying compensation to a loan originator that is based on a term of a transaction, the terms of multiple transactions by an individual loan originator, or a proxy for the terms. The rule also prohibits a loan originator from receiving compensation based on those terms, and it prohibits the loan originator from receiving compensation from both the consumer and any other person on the same transaction (the "dual compensation" prohibition at 1026.36(d)(2)).

The rule was drafted to address the pre-crisis pattern of mortgage brokers being paid larger commissions for higher-rate loans, which produced a steering incentive against the consumer. The rule's language reaches beyond the mortgage broker scenario, and its definitions of "loan originator" and "compensation" have been read broadly by the CFPB in supervisory findings and by state attorneys general in enforcement actions. The rule's operational bite on AI vendor relationships is not obvious at contract signature. It becomes obvious when the vendor's pricing structure or the vendor's operational role brings the vendor's economic outcomes inside the rule's perimeter.

We build the AI agent that participates in mortgage-origination intake and application processing at lenders, brokers, and depository institutions. The rule matters to us because our customers ask, on average once per procurement cycle, whether the fees they pay us for AI-agent activity are compensation to a loan originator under Reg Z 1026.36. The right answer requires close reading of the rule's definitions, the CFPB's commentary, and the specific facts of the arrangement, and we have run the analysis often enough to produce the structure below.

## Who Is a Loan Originator Under 1026.36(a)(1)

The rule at [1026.36(a)(1)](https://www.consumerfinance.gov/rules-policy/regulations/1026/36/) defines "loan originator" as "a person who, in expectation of direct or indirect compensation or other monetary gain or for direct or indirect compensation or other monetary gain, performs any of the following activities: takes an application, offers, arranges, assists a consumer in obtaining or applying to obtain, negotiates, or otherwise obtains or makes an extension of consumer credit for another person; or through advertising or other means of communication represents to the public that such person can or will perform any of these activities." The definition covers both the individual and the organization: the "individual loan originator" is the natural person, and the "loan originator organization" is the entity.

The rule at 1026.36(a)(1)(ii) explicitly excludes "an individual who does not take a consumer credit application or offer or negotiate credit terms available from a creditor," provided the individual only assists a consumer by advertising, referring the consumer to a specific creditor or loan originator, or providing information the consumer requested that describes a specific transaction. The "loan processor" exclusion at 1026.36(a)(1)(iv) further clarifies that individuals who perform "clerical" or "administrative" work under the supervision of an actual loan originator are not themselves loan originators for purposes of the rule.

The AI vendor's status under this definition is the analytical starting point. If the AI agent takes applications, offers or negotiates terms, or assists the consumer in obtaining credit, the vendor's employees who supervise or operate the agent may fall within the individual-loan-originator definition, and the vendor's organization may fall within the loan-originator-organization definition. The exclusions do not automatically apply to AI vendors, because the rule's "clerical or administrative" exclusion assumes the individual is working "under the supervision" of an actual loan originator at the creditor, and the AI vendor is typically not organized that way.

Our position, worked through with counsel for our customers, is that the AI agent is a tool used by the creditor's or broker's own loan originators, that the creditor's or broker's own loan originators are the "loan originators" under the rule, and that the AI vendor's role is that of a technology provider whose employees are not themselves loan originators because they are not participating in the specific transactions. The position depends on the specific contract structure and operational allocation of activities, and it is not an automatic conclusion; a vendor whose employees are personally engaged in negotiating loan terms with consumers for specific transactions is a vendor whose employees are loan originators.

## What Counts as "Compensation" Under 1026.36(a)(3)

The rule's definition of "compensation" at [1026.36(a)(3)](https://www.consumerfinance.gov/rules-policy/regulations/1026/36/) is broad: "salaries, commissions, and any financial or similar incentive." The [official commentary at 1026.36(d)(1)-1](https://www.consumerfinance.gov/rules-policy/regulations/1026/Interp-36/) makes clear that compensation can be from the creditor, from an affiliate of the creditor, or from any other person, and it can be paid directly or indirectly.

The scope of "compensation" is where the AI vendor's contract terms have to be looked at closely. A vendor fee paid by the creditor to the vendor for the vendor's software service is not compensation to a loan originator if the vendor is not itself a loan originator. If the vendor's fee is per-loan-originated, the fee is a compensation-adjacent payment, and the question is whether the fee is compensation to a person who is a loan originator or to a person who is not.

The CFPB's [2013 final rule on LO Comp](https://www.consumerfinance.gov/policy-compliance/rulemaking/final-rules/loan-originator-compensation-requirements-under-truth-lending-act-regulation-z-2013/) and the subsequent amendments have specifically addressed several third-party arrangements. Lead-generation payments to consumer-facing referral entities can be structured as compensation to a loan originator if the referral entity is itself a loan originator under the definition. Technology-license fees paid to loan-origination-system vendors have historically not been treated as loan-originator compensation, because the LOS vendor is not itself a loan originator.

The AI vendor sits closer to the LOS-vendor position than to the referral-entity position for typical arrangements. The vendor is providing technology; the creditor's own loan originators are using the technology; the vendor's fee is for the software and service, not for the origination outcome per se. The analysis holds if the vendor's contract and operational role support the technology-provider characterization.

## The Terms-Based Compensation Prohibition and the Per-Loan Fee Question

The rule at [1026.36(d)(1)](https://www.consumerfinance.gov/rules-policy/regulations/1026/36/) prohibits any person from paying to a loan originator "compensation in an amount that is based on a term of a transaction," a term of multiple transactions by the same originator, or a proxy for the term. Terms include the interest rate, the loan program, the loan amount within certain limits, and other rate-and-cost variables. Proxies include factors that consistently vary with a term and that are not defined by other rule-permitted factors.

The rule at 1026.36(d)(1)(iii) permits compensation to be based on the loan amount, provided the compensation is a fixed percentage of the amount (or a fixed amount plus a fixed percentage). Compensation can also be based on the loan originator's overall dollar volume, the number of loans originated over a period, and long-term performance metrics measured over at least 12 months, subject to specific rule-defined limits.

An AI vendor's per-loan fee raises the terms-based question, because the per-loan fee is by definition based on the fact that a loan was originated. The question is whether the per-loan fee to the vendor is treated as compensation to a loan originator. If the vendor is not a loan originator, the fee is not LO compensation and 1026.36(d)(1) does not apply. If the vendor is a loan originator (because, for example, the vendor's employees actually negotiate loan terms with consumers), the fee's structure has to comply with the rule, which for a per-loan structure would mean the fee has to be a fixed dollar amount per loan or a fixed percentage of the loan amount.

The vendor structure we use with our customers is a subscription plus usage arrangement where the subscription is a fixed fee independent of originated volume and the usage component is per-conversation or per-application, not per-originated-loan. The distinction between a per-application fee and a per-originated-loan fee is meaningful under the rule: the per-application fee is for the technology usage regardless of outcome; the per-originated-loan fee tracks the origination outcome and is closer to a compensation arrangement. The per-application-vs-per-origination distinction is one we discuss with our customers' legal teams at contract review.

## The Dual-Compensation Prohibition and the Consumer-Paid Fee Question

The rule at 1026.36(d)(2) prohibits a loan originator from receiving compensation directly from a consumer in a transaction where the loan originator (or the loan originator's employer) also receives compensation from any other person in connection with the transaction. The dual-compensation prohibition means that a broker whose consumer pays the broker's fee cannot also receive a yield-spread premium or lender-paid compensation on the same transaction.

The AI vendor's fee typically is not paid by the consumer. The creditor pays the vendor for the technology usage; the consumer does not directly compensate the vendor for the AI agent's participation. The fee flow is creditor-to-vendor rather than consumer-to-vendor. The dual-compensation analysis does not typically implicate the vendor's fee, because the fee is not from the consumer.

Where the analysis becomes closer is if the AI vendor's fee is embedded in a fee the consumer pays. A creditor whose application-fee schedule includes a specific charge that is directly passed through to the AI vendor is a creditor whose consumer-paid fee is functionally paying the vendor. The pass-through structure is not intrinsically problematic under 1026.36(d)(2) because the vendor is typically not a loan originator, but the arrangement raises the fee-disclosure question under [TILA 1026.4](https://www.consumerfinance.gov/rules-policy/regulations/1026/4/) and the fee's characterization for QM points-and-fees purposes under 1026.32(b)(1). The pass-through structure is one we advise our customers against, because it complicates the fee-disclosure posture without providing any operational or economic advantage.

## The Steering and Anti-Steering Rules and the Agent's Program Choice

The rule at 1026.36(e) sets the anti-steering safe harbor for loan originators. The safe harbor requires the loan originator to present the consumer with loan options from a significant number of the creditors with which the loan originator regularly does business, and the options have to include specific rate-and-cost combinations from those creditors. The safe-harbor mechanics are for the loan originator's own compliance and do not directly apply to the AI vendor, but the AI agent's operational role can implicate the loan originator's safe-harbor posture.

If the AI agent recommends a specific loan program or a specific creditor to the consumer, the recommendation is the loan originator's recommendation for anti-steering purposes, and the loan originator's safe harbor requires the anti-steering options to be presented. The agent's program-selection logic has to align with the loan originator's anti-steering compliance approach: either the agent presents the full set of safe-harbor options and lets the consumer choose, or the agent's recommendation is documented as the loan originator's recommendation with the anti-steering options separately available.

The agent's role in the anti-steering compliance is operational: the agent produces the safe-harbor option set, presents it to the consumer, and records the consumer's choice. The loan originator's compliance file includes the specific options presented and the consumer's choice, and the file's audit trail supports the safe-harbor claim.

## The Compensation-Attribution Analysis for the AI Vendor's Own Compensation

The rule's prohibition on terms-based compensation applies to compensation received by loan originators. If the AI vendor's employees who supervise the agent are not loan originators themselves, the compensation those employees receive is not subject to the rule. Our analysis for vendor employees is that our engineers, product managers, and account managers are not loan originators, because they do not perform any of the activities in 1026.36(a)(1) with respect to specific consumer credit transactions. The employees build and operate a software product; they do not take applications, offer or negotiate credit terms, or assist specific consumers in obtaining specific loans.

The analysis would be different for a vendor whose employees are personally engaged with specific consumers on specific transactions. A vendor whose account managers help specific consumers work through specific loan applications, negotiate specific terms with specific creditors, or make recommendations on specific programs would have employees who fall within the loan-originator definition, and those employees' compensation would be subject to the rule.

The distinction between the vendor whose employees are personally engaged and the vendor whose employees are operating the technology is the operational anchor for the analysis. Our operational model keeps our employees out of the specific-transaction perimeter, and the model's discipline is not just a business preference; it is a compliance boundary that keeps the vendor's employees out of the rule's coverage.

## The Contract Provisions That Anchor the Analysis

The vendor contract we use with our customers has specific provisions that anchor the LO Comp analysis. The contract identifies the vendor as a technology provider and not a loan originator, states that the vendor's employees will not personally engage with specific consumers on specific transactions, and allocates responsibility for LO Comp compliance to the customer. The contract's fee structure is subscription-plus-usage as described earlier, with the usage component measured on inputs (conversations, applications received) rather than outputs (loans originated).

The contract also includes provisions on the customer's own loan-originator compensation program: the customer represents that the customer's loan-originator compensation complies with 1026.36(d), that the customer's use of the AI agent does not itself produce compensation to any loan originator based on transaction terms, and that the customer will notify the vendor if the customer's use case shifts to a structure that would implicate LO Comp for the vendor's role.

The contract is not the entirety of the compliance posture, but it is the artifact the CFPB or the state examiner will read first when the question is asked. A contract that clearly allocates the LO Comp analysis and that structures the fee to be independent of the transaction terms is a contract whose position on the rule is defensible. A contract whose fee is per-originated-loan and whose vendor role is ambiguous is a contract whose position is at risk.

## The Recordkeeping Requirements at 1026.25(c)(2) and What the Vendor Has to Preserve

The rule at [1026.25(c)(2)](https://www.consumerfinance.gov/rules-policy/regulations/1026/25/) requires the creditor to maintain records evidencing compliance with 1026.36(d) and (e) for at least three years after the date of the transaction. The records include compensation agreements, records of compensation actually paid, and records evidencing the anti-steering compliance if the loan originator relied on the safe harbor.

The AI vendor's records support the creditor's recordkeeping obligation to the extent the vendor's activity is documented in the creditor's file. Our records for our customers include the specific agent-consumer conversations, the specific options presented to the consumer, the specific consumer choices, and the specific escalations to the creditor's own loan originators. The records are available to the creditor for the creditor's own recordkeeping and for any examination or enforcement inquiry.

The vendor's own retention posture on the records is a matter of the contract and the data-processing arrangement. The contract typically requires the vendor to retain the records for the creditor's retention period and to produce the records on request. The vendor's compliance with the retention arrangement is a specific operational discipline the vendor's own compliance function manages.

## The State-Specific Overlays and the Higher-Bar Question

The federal LO Comp rule is a floor; several states have overlays that impose additional requirements on loan-originator compensation and on compensation structures more broadly. California's [Financial Code Section 4995](https://leginfo.legislature.ca.gov/faces/codes_displaySection.xhtml?lawCode=FIN&sectionNum=4995.), New York's [Banking Law Article 12-E](https://www.dfs.ny.gov/) provisions, and specific state-by-state licensing regulations for mortgage brokers impose compensation-structure requirements that overlap the federal rule.

The vendor's contract with the customer identifies the states the customer operates in, and the vendor's operational model is consistent with the federal floor. The customer's own compliance program addresses the state overlays as part of the customer's licensing and supervision posture. The vendor is not typically responsible for the state overlays, but the vendor's operational conduct that respects the federal floor is generally consistent with the state overlays as well.

## The CFPB Circular on Consumer-Facing AI and What It Signals for LO Comp

The CFPB has issued [Circular 2023-03 on unfair, deceptive, or abusive practices arising from chatbots](https://www.consumerfinance.gov/compliance/circulars/circular-2023-03/) and has indicated broader interest in consumer-facing AI in financial services. The Circular addresses UDAAP rather than LO Comp specifically, but it signals a supervisory interest in the AI vendor's role that will eventually extend to the rules the vendor's activity intersects.

The vendor whose LO Comp analysis is clean before the supervisory attention arrives is the vendor whose analysis will hold up when the attention arrives. The vendor whose analysis is ambiguous is the vendor whose analysis will be tested at the moment the vendor's customer is under supervisory review. The right time to structure the analysis is at contract signature; the wrong time is at examination.

## The Failure Mode We Engineer Against

The pattern that produces the worst LO Comp outcomes for AI vendors is the vendor whose pricing model was structured for commercial simplicity (per-originated-loan pricing that is easy to explain to procurement teams) without the rule's constraints in mind, whose employees have gradually taken on transaction-specific engagement to help customers with particular issues, and whose contract does not clearly allocate the LO Comp analysis. The vendor's arrangements produce operational patterns that read as loan-originator activity, and the CFPB's or the state's inquiry into the vendor is a matter of when rather than if.

The architecture we run against this is that the vendor's pricing is structured for the rule, the vendor's operational model is disciplined about the vendor-employee-vs-transaction perimeter, and the contract with the customer allocates the analysis explicitly. The discipline is not just a compliance choice; it is a business choice that keeps the vendor's arrangements simple and the vendor's exposure to LO Comp inquiries low. The vendor whose pricing is per-application rather than per-origination and whose employees are technology operators rather than transaction participants is a vendor whose LO Comp posture is clear.

## The Honest Read

The Loan Originator Compensation rule at Reg Z 1026.36 is one of the most operationally sensitive rules in mortgage origination, and its reach extends into the AI vendor relationship in ways the contract's commercial terms may not anticipate. The analysis is not automatically simple, and the wrong contract structure can bring the vendor's fees and the vendor's employees inside the rule's perimeter without the parties realizing until an examination raises the question.

The AI vendor whose analysis is clean is the vendor whose customer's compliance posture is not at risk from the vendor's own arrangements. The customer's own LO Comp program is complex enough without the vendor's fees adding a new dimension, and the vendor's job is to be a clean operational and commercial partner rather than a source of compliance ambiguity. The discipline we run internally on this is one of the specific operational advantages our customers value, and it is one of the specific things that make an AI vendor operationally trustworthy in regulated finance.

We have written separately on the [SAFE Act and MLO licensing analysis for AI mortgage assistants](/blog/safe-act-ai-mortgage-assistants-nmls-licensing-line), on the [third-party risk management framework for AI vendors](/blog/third-party-risk-management-ai-vendors-banking-tprm-playbook), on the [Reg Z 1026.43 ATR/QM rules](/blog/reg-z-1026-43-ability-to-repay-qualified-mortgage-ai-underwriting), and on the [adverse-action notice rules under ECOA/Reg B](/blog/adverse-action-notices-ai-credit-decisions-ecoa-reg-b). The LO Comp rule sits alongside these as part of the origination program's compliance architecture, and the AI vendor whose analysis on each is clean is the vendor whose customer's origination program is stronger for the partnership.

---

_Source: [https://www.seiright.com/blog/reg-z-1026-36-loan-originator-compensation-ai-vendor-pricing](https://www.seiright.com/blog/reg-z-1026-36-loan-originator-compensation-ai-vendor-pricing) · Sei AI_