Regulatory relief that lapses at the next release

The CFPB’s December 2018 sandbox proposal tied every letter, approval and exemption to a described product. That made relief a matter of change control, and product changes later ended two of the best-known grants.

Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.

Two-minute summaryThe dek and the first sentence of each section, quoted from the article

The CFPB’s December 2018 sandbox proposal tied every letter, approval and exemption to a described product. That made relief a matter of change control, and product changes later ended two of the best-known grants.

  1. The week On December 13, 2018, the Federal Register published a proposal from the consumer bureau created by the Dodd-Frank Act, which at the time styled itself the Bureau of Consumer Financial Protection, or BCFP.
  2. How the relief was built The proposal offered three instruments, and they differed sharply in legal force.
  3. What the relief attached to The proposal used one phrase throughout: the described aspects of the product or service.
  4. What could end the relief The proposal expected each no-action letter to state its revocation grounds, and it listed the ones it anticipated.
  5. What a holder had to keep producing Part I dropped the data-sharing commitment of the 2016 policy.
  6. Who could hold the relief The proposal invited applications from trade associations, service providers and other third parties.
  7. The strongest objections One objection is that very few firms ever held relief, so its design affected almost nobody.
  8. How it differs by type of firm A large bank has mature change management, model risk management and third-party oversight.
  9. With hindsight The Bureau finalized the December 2018 proposal in two documents, published September 13, 2019 and applicable from September 10, 2019.

The week

On December 13, 2018, the Federal Register published a proposal from the consumer bureau created by the Dodd-Frank Act, which at the time styled itself the Bureau of Consumer Financial Protection, or BCFP. The document was titled Policy on No-Action Letters and the BCFP Product Sandbox. It came from the Bureau’s Office of Innovation. Written comments were due by February 11, 2019.

The proposal started from a candid admission. The Bureau had proposed its first no-action letter policy in October 2014 and finalized it in February 2016. Under that 2016 policy, a no-action letter was a non-binding recommendation from staff. By December 2018 the Bureau had issued only one such letter. It said that record strongly suggested the process, and the relief on offer, had not given firms enough reason to apply.

The answer had two parts. Part I rewrote the no-action letter itself. Letters would be issued by authorized officials of the Bureau rather than as staff recommendations, so the Bureau itself would stand behind them. Applicants would no longer have to commit to sharing data. Letters would have no time limit by default. Letters about unfair, deceptive or abusive acts or practices would no longer be expected to be rare. The Bureau said it expected to grant or deny an application within 60 days of deeming it complete.

Part II created the BCFP Product Sandbox. Firms admitted to it could receive three kinds of relief. The first was no-action relief like that in Part I. The second was an approval under the safe harbor provisions of three statutes: the Truth in Lending Act, the Equal Credit Opportunity Act and the Electronic Fund Transfer Act. The proposal said a recipient acting in good faith in conformity with an approval would be immune from enforcement by any federal or state authority, and from private lawsuits. The third was an exemption by order from certain statutory or regulatory provisions.

Approvals and exemptions would be time-limited. The Bureau expected two years to be appropriate in most cases, with a route to extension where there was evidence of consumer benefit and no consumer harm. Sandbox applicants would have to commit to sharing aggregate data on the product’s effect on consumers, on a proposed schedule. They would also describe consumer risks, plans for unanticipated harms and the resources available for restitution.

Every form of relief was tied to what the proposal called the described aspects of the product or service. That phrase covered both the product and the particular manner in which it was offered. The Bureau gave an example. If a letter covered only written disclosures, oral marketing by call center staff could still face supervisory or enforcement action.

The proposal invited applications from trade associations and service providers as well as from the firms offering products. It also set out the grounds on which relief could be revoked, and promised to publish letters on the Bureau’s website. The Bureau characterized the whole policy as a general statement of policy and a procedural rule, exempt from notice-and-comment requirements, and asked for comment anyway.

The December 2018 sandbox proposal is usually remembered as a regulatory story. A consumer protection agency offered fintech firms a softer path to market. Consumer groups objected, industry applauded, and the policies that came out of it were rescinded in 2022. That account is accurate as far as it goes. It misses what the proposal meant for the people who build and run consumer financial products.

Our argument is that the proposal made regulatory relief a property of a product’s configuration. Every letter, approval and exemption attached to the described aspects of a product, as offered in a described manner. Relief therefore lasted only as long as the product stayed inside its description. That turned a legal benefit into an engineering and change control commitment, and it put technology and product leaders in charge of whether the relief held.

We test that argument through four parts, after a short account of how the relief worked: what relief attached to, what could end it, what a holder had to keep producing, and who could hold it when banks and their service providers build products together. A sixth section weighs the strongest objections. The last looks at how the argument plays out for different kinds of firm.

The sector is financial services, because every tool in the proposal operated under federal consumer financial law. The documents cited are policy statements and orders, not legislative rules. Firms considering any form of regulatory relief should work through their specific position with counsel.

How the relief was built

Three instruments with different strength

The proposal offered three instruments, and they differed sharply in legal force. A no-action letter was an exercise of the Bureau’s supervisory and enforcement discretion. It said that, subject to good faith and substantial compliance with its terms, the Bureau would not make supervisory findings or bring an action over the described product. It bound the Bureau’s discretion and nothing else.

An approval was stronger. Three statutes, the Truth in Lending Act, the Equal Credit Opportunity Act and the Electronic Fund Transfer Act, each contain a safe harbor for acts done in good faith in conformity with an approval by the Bureau. The proposal read those provisions to mean that a recipient would be immune from enforcement by federal and state authorities and from lawsuits by private parties.

An exemption by order removed the obligation itself, either under statutory exemption powers or under the Bureau’s rulemaking authority for regulatory provisions that did not mirror a statute. The proposal said that an exemption, too, would give immunity from federal and state enforcement and from private suits based on the exempted provisions.

Why the strength mattered to builders

For a product team, the difference between these instruments was the difference between a supervisory understanding and a legal defense. A no-action letter told the firm the Bureau would not come after it. It said nothing about state attorneys general, state regulators or plaintiffs. An approval reached all three, but only for the three statutes named and only for conduct in conformity with the approval.

So the first design question for any applicant was which statutes created the uncertainty. A new disclosure flow for an open-end credit product sat within the Truth in Lending Act. A credit model that used new data sat within the Equal Credit Opportunity Act. A new way of moving money between accounts could sit within the Electronic Fund Transfer Act. Problems under other laws, or under the general prohibition on unfair, deceptive or abusive practices, could reach only no-action relief.

Who signed and how fast

Under the 2016 policy, a letter was a staff recommendation. The proposal moved signature to authorized officials of the Bureau, signed by the Assistant Director of the Office of Innovation or other authorized members of that office. The Bureau said the change would give recipients greater assurance that the Bureau itself stood behind the relief. For a firm deciding whether to invest in a product, that assurance was the point of applying.

The proposal also set a target. The Bureau intended to grant or deny an application within 60 days of notifying the applicant that it had deemed the application complete. It intended to respond to reasonable requests to reconsider a denial within 60 days as well. The clock started at completeness, not submission. A firm planning a launch around relief would need to count the time spent answering the Bureau’s questions before the clock began.

Applicants were encouraged to contact the Office of Innovation informally before applying. That informal stage matters for planning. A product team that waits for legal clearance before finishing its design may find the description it filed no longer matches what it built. A team that finishes the design first may find the Bureau wants changes. Either way, the description and the build have to converge before the relief has value.

A policy, not a rule

The proposal described itself as a general statement of policy and a procedural rule. It said the policy would not impose legal requirements on third parties or create rights enforceable by them. The individual letter, approval or exemption was where the legal effect sat. That distinction mattered later, because a policy statement can be withdrawn without rulemaking. Firms that relied on the program were relying on two things: the instrument they held, and the agency’s continued willingness to run the program that issued it.

What the relief attached to

The described aspects of the product

The proposal used one phrase throughout: the described aspects of the product or service. A footnote defined it to cover both the particular aspects of the product and the particular manner in which it was offered or provided. Each letter was to state that it applied only to those aspects, and not to different aspects of the same product.

The Bureau’s own illustration showed how narrow that could be. A letter that covered written disclosures did not cover oral marketing by call center representatives. The same product, sold through a different channel, sat outside the relief. For a firm, the boundary of protection was the boundary of its application.

What an application had to describe

The application elements show what the Bureau expected to see. For a no-action letter, an applicant described how the product functioned, the terms on which it would be offered, and the manner of offering, including consumer disclosures. It explained the potential consumer benefits and risks and how it would mitigate the risks. It identified the statutory and regulatory provisions at issue and the uncertainty, ambiguity or barrier that relief would address.

A sandbox application added more. It asked for the requested duration and any limits on transaction volume, number of consumers or geography. It asked for suggested metrics, such as consumer utilization, for judging whether the benefits were realized. It asked for plans for unanticipated harms and the resources available for restitution.

Why a description becomes a specification

Read as an engineer would read it, that application was a product specification. It fixed the functional behavior, the terms, the channel, the disclosure text and, for the sandbox, the scale. The relief attached to that specification. Any change that moved the product outside it moved the product outside the relief.

Consumer financial products change often. Pricing tables are adjusted. Underwriting models are retrained. Disclosure screens are redesigned after usability testing. A mobile release can change the order in which a consumer sees terms. Under the proposal’s structure, each of those changes forced the same check against the product on file.

Products built on models

The configuration problem is sharpest for products whose behavior is set by a model rather than by fixed rules. A credit model can change its outputs without any change to the code around it, simply by being retrained on newer data. A firm may add an attribute, drop one, or change how missing values are handled. Each change can alter who is approved and on what terms.

Relief granted for one model therefore described something that would not stay still. A firm would need to decide which model changes counted as changes to the described aspects. A retraining on the same attributes with the same method might stay inside the description. A new attribute or a new method probably would not.

Those are judgments, and the proposal left them to the holder. That made the model governance process, already familiar to banks, the place where relief was kept or lost. The documentation a bank produces when it validates a model change is close to what it would need to show that a change stayed inside the relief.

A simple change classification

A firm holding relief could sort proposed changes into three groups before release:

  • Outside the description. Changes that do not touch the functions, terms, channel or disclosures on file, such as infrastructure upgrades or internal tooling. These proceed under normal change control.
  • Inside the description, within tolerance. Changes the application anticipated, such as periodic retraining on the same attributes, recorded with evidence that they match the filed method.
  • Inside the description, beyond tolerance. Changes to the described aspects themselves, such as a new fee, a new data source or a new channel. These go to the Bureau as an amendment, or proceed only after the firm accepts that the relief will no longer cover them.

The value of the scheme depends on the second group. The more precisely the application described expected change, the more changes could proceed without amendment. A firm that wrote its application around a snapshot of the product left itself no tolerance at all.

What a firm would build to answer it

The answer had to come from systems, not from lawyers’ memories. A firm holding relief would need a controlled record of the described product, with a version number. It would need a change process that flagged any change touching the described aspects. It would need someone with authority to decide whether a change stayed within the description, required an amendment, or required the firm to accept that the relief no longer applied.

For a credit product, that record would include the model version and the data attributes it used, the decision rules, the adverse action reasons and the disclosure templates. For a payments product, it would include the transaction flows, the error resolution process and the fee schedule. In each case, the evidence a regulator might later request was evidence that the product in market matched the product on file.

What could end the relief

Revocation grounds

The proposal expected each no-action letter to state its revocation grounds, and it listed the ones it anticipated. The first was failure to comply substantially and in good faith with the letter’s terms. The second was a Bureau determination that the product was causing material, tangible harm to consumers. The third was a change in the legal basis for the letter through a statutory change or a Supreme Court decision.

The process was protective. The Bureau would notify the recipient of the grounds and allow a reasonable time to respond. Where the ground was a compliance failure, it would offer a chance to cure. Revocation would be written and reasoned. The Bureau intended to allow a wind-down period, unless the product was causing material harm and a wind-down would make it worse. The Bureau also said, citing other agencies’ programs, that it expected revocation to be quite rare.

Changes in law

The third ground deserves a closer look. The proposal limited it to a change in the legal basis for the relief through a statute or a Supreme Court decision. Lower court decisions, new Bureau rules and changes in Bureau leadership were not listed. The Bureau said it was particularly interested in comment on whether other changes in law should be added.

For a holder, that ground creates a monitoring duty. Someone must track the statutes and Supreme Court cases that bear on the relief and assess whether a change removes its basis. That is ordinary regulatory change management, applied to a specific instrument. It is easy to forget once the letter is framed and the product is live.

Time limits

No-action letters under Part I would have no time limit by default. Approvals and exemptions under the sandbox would. The Bureau expected two years to suit most cases and set out an extension route. It said it intended to grant extensions where there was evidence of consumer benefit and an absence of consumer harm.

That made the data commitment, discussed below, the key to the extension. A firm that wanted its safe harbor to last beyond two years had to show evidence. The evidence had to come from the product’s own operation, collected from the start.

The ground the proposal did not list

None of the listed revocation grounds covered the most likely way for relief to end. A firm could simply change its product. Because relief applied only to the described aspects, a product that drifted outside its description lost protection without any act by the Bureau. No notice, no cure period and no wind-down applied, because nothing had been revoked. The firm had stepped outside the relief on its own.

That gap is the center of our argument. The revocation section looked like the main risk to a holder. For a fast-moving product, the main risk was its own roadmap. A firm that released weekly faced the question of whether it was still covered with every release.

What a holder would do about it

Three practical responses follow from the structure. The first is to apply for relief at the level of the stable design decision, not the volatile implementation detail. A letter about the category of data a model may use is more durable than a letter about one model version.

The second is to plan for amendment. The proposal said applicants could withdraw, modify and resubmit applications at any time. A firm with a planned change could bring it to the Bureau before release rather than after. That requires a release calendar that includes regulatory lead time, which product calendars do not always include.

The third is to decide, in advance, which changes the firm would forgo to keep the relief. That is a business decision with a cost, and it belongs with the executive who owns the product, not with the compliance function alone.

What a holder had to keep producing

The data commitment

Part I dropped the data-sharing commitment of the 2016 policy. The Bureau said it had been redundant or unduly burdensome, and noted that commenters on the 2016 policy had called the volume of required information excessive. Part II kept a data commitment for the sandbox. Applicants had to describe the data they held or would develop on the product’s impact on consumers, and propose a schedule for sharing it. A footnote said the data should be aggregate data.

Aggregate data still has to be produced from somewhere. A firm would need to define the measures, build the pipelines that compute them from transaction and customer records, and keep the definitions stable over time so that one period could be compared with the next. If the measures were the same ones the firm had proposed as evidence of consumer benefit, they would also be the measures that decided its extension.

Restitution and harm plans

Sandbox applications were to explain plans for unanticipated consumer harms and the resources available for restitution for material, quantifiable economic harm. That element is unusual for a relief program. It asked a firm to plan, before launch, how it would identify affected consumers, calculate their losses and pay them.

Each of those steps depends on data the firm may not keep by default. Identifying affected consumers needs a reliable record of who received which version of the product. Calculating loss needs the terms each consumer actually received. Paying them needs current contact and account details. A firm that could not reconstruct those facts could not credibly promise restitution.

Confidentiality and publication

The proposal said the Bureau intended to publish no-action letters on its website and, in appropriate cases, a version or summary of the application. It might publish denials too, after giving the applicant a chance to seek reconsideration. Applicants that wanted confidential treatment were to ask in a separate letter, citing the Freedom of Information Act, the Bureau’s disclosure rule or other law. Applications were not to include personally identifiable information, except as the policy provided.

For a technology leader, that set two tasks. One was to separate what could be published from what was proprietary, such as model features or fraud controls. The other was to make sure nothing in the application or the data commitment carried customer-level personal information unless it was required and protected.

What an examiner would ask to see

A holder should expect its relief to come up in supervision. The proposal said the Bureau kept the right to obtain information about a product covered by a letter under its normal supervision and enforcement powers. The evidence a firm would want ready includes:

  • the application as filed, the letter or order as issued, and every amendment;
  • a record of the product in market at each point in time, by version, with the dates each version was live;
  • the change log for the described aspects, with the classification and approval for each change;
  • the data delivered to the Bureau under any data commitment, with the definitions and code that produced it;
  • consumer complaints about the product, and how each was resolved.

Much of this exists in a well-run firm already. Linking it to a relief instrument is the new work.

The cost of producing it

None of this was free. Ongoing data reporting, harm monitoring and restitution planning required people and systems for the life of the relief. Firms that treated relief as a one-time legal win would find that it came with an operating cost. Firms that already ran disciplined product analytics would find much of it was work they did anyway, pointed at a new audience.

Who could hold the relief

Third-party applicants

The proposal invited applications from trade associations, service providers and other third parties. A trade association could apply on behalf of members. A service provider, defined by reference to the Dodd-Frank Act as a person providing a material service to a covered person, could apply together with existing or prospective clients.

The Bureau acknowledged a chicken-and-egg problem. Some service providers could not apply without a client relationship, and could not get the relationship without relief. Its answer was a provisional letter, issued to a third party that could not submit a complete application, followed by a non-provisional letter once the gaps were filled. Other entities described in the application could receive the letter at the same time or later by notifying the Bureau and supplying the needed information.

Templates

The third-party route made a further form of relief possible: a template that other firms could adopt. The Bureau’s list of granted applications shows several. It records a no-action letter and a template for the Department of Housing and Urban Development, on behalf of HUD-approved housing counseling agencies. It also records no-action letter templates for Brace Software and for the Bank Policy Institute, and an approval template for Build Commonwealth.

A template shifts the configuration problem. The template describes the product, and each adopting firm must keep its own product inside that description. An adopter that customizes the product to fit its own systems can move outside the template without noticing. Adoption therefore needs the same change control as an individual letter, applied to a description the adopter did not write.

Why that mattered for banks and their providers

Many consumer products are assembled by more than one firm. A bank may hold the charter and the credit, a technology firm may run the application flow and the model, and a processor may move the money. Each party controls some of the described aspects. None controls all of them.

That arrangement complicates the configuration problem. If a service provider holds the relief and the bank makes a change to its pricing, the product may leave its description. If the bank holds the relief and the provider retrains a model, the same thing can happen. Relief that depends on the product staying inside its description depends on every party that can change the product.

What a contract would need to say

A bank and a provider relying on shared relief would need contract terms that track the relief. Those include a list of the described aspects each party controls, notice before any change to them, and a joint decision on whether a change needs an amendment. They also include allocation of the data commitment and the restitution plan, and a defined exit if the relief ends.

For a bank, the relief structure added one more item to third-party oversight: whether a provider’s change could cost the bank a regulatory protection it had relied on. Third-party risk teams would not ordinarily look for that.

Planning the exit

Every form of relief in the proposal could end: by revocation, by expiry of a sandbox term, by a change in law, or by the product leaving its description. A product that depends on relief should therefore have an exit plan written before launch. The plan would say what the product does if relief ends, which features are switched off, which disclosures change and how existing customers are treated.

The proposal’s wind-down provision helps only when the Bureau revokes. It offers nothing when the holder changes the product. In that case the firm sets its own transition, and it needs to know in advance whether the changed product complies without relief. If it does not, the change should not ship.

Coordination with other regulators

The proposal cited the Dodd-Frank Act’s instructions to coordinate with federal agencies, state regulators and state attorneys general. It said the Bureau was interested in agreements with state authorities that offered similar relief, and that applicants could ask the Bureau to coordinate with named regulators. It warned that extensive coordination could slow its response. A firm operating in many states would have to decide whether to seek relief from each regulator separately and how to keep the descriptions consistent.

The strongest objections

Relief was rare, so the problem was small

One objection is that very few firms ever held relief, so its design affected almost nobody. The Bureau’s own list of granted applications supports the premise. As checked on November 14, 2026, it named six recipients of no-action letters or templates and three recipients of approvals or approval templates under its innovation policies.

The objection misses how relief was used. Several recipients were templates or trade groups, including a template sought by the Bank Policy Institute. Templates allowed other firms to obtain relief on the same terms. And the configuration problem did not depend on volume. For each holder, the relief was only as good as its change control, and the two best-documented terminations both turned on product changes.

Lawyers, not engineers, owned the relief

A second objection is that relief was a legal instrument, negotiated and monitored by counsel. That is true of the negotiation. It is not true of the monitoring. Counsel cannot see every model retraining or fee table change. Only the systems that manage those changes can flag them. The relief’s durability depended on the firm’s engineering processes, whoever signed the application.

Consumer groups were right, and the program was the problem

A third objection comes from the other side. Consumer groups argued that the proposal was a step backward from the 2016 policy, and one of two groups of state attorneys general agreed. On that view, the structure of relief matters less than whether it should have existed.

We take no position on that policy question. Our argument holds either way. If relief exists, it attaches to a described product, and the holder’s change control decides its value. The January 2025 version of the program, issued by the same agency that had rescinded the 2019 policies, kept that feature and made it automatic, before the Bureau withdrew it in May 2025.

Relief invited risk-taking

A fourth objection is that relief encouraged firms to take risks they would otherwise avoid, and that change control is a secondary issue. The proposal’s structure partly answers this. Sandbox approvals were time-limited, required data sharing and asked for restitution plans. The revocation grounds included material, tangible harm to consumers. Whether those safeguards were enough is a policy judgment. For the firm, they added to the operating commitments that came with relief, which strengthens the case that relief had to be run as well as held.

Where the argument is weaker

The argument is weakest for no-action letters about stable, narrow questions. A letter about the form of one disclosure, for a product that rarely changes, needs little change control. It is strongest for products built on models, pricing engines or fast release cycles, which describes many of the fintech products the program was designed to encourage.

How it differs by type of firm

Large banks

A large bank has mature change management, model risk management and third-party oversight. It also has many products and many teams that can change them. Its challenge is linking a relief instrument to the specific systems in scope, so that a change in one team’s release does not breach the description held by another. The Bureau’s list of granted applications included letters to two large national banks, Bank of America and JPMorgan Chase.

The practical step is to tag the systems in scope. A bank’s configuration management database or application inventory can carry a flag on every system that implements a described aspect. Change tickets against a flagged system then route to the owner of the relief for review. The same flag lets internal audit sample changes and test whether the review happened. Without it, the bank depends on individual teams remembering an instrument they may never have read.

Community banks and credit unions

Smaller institutions often rely on core processors and other providers for the systems in question. They may control few of the described aspects themselves. For them, relief obtained through a trade association or a provider’s template could be more practical than an individual application. The contract terms described above matter more, because the institution depends on the provider to keep the product inside its description.

Fintechs and service providers

Fintechs and providers tend to release often and change models frequently. They are the firms the program was built for and the firms most exposed to the configuration problem. A provider holding relief for many clients would need to manage the description across all of them, and would need clients to notify it of their own changes.

Nonbank lenders and payment firms

Nonbank firms under state licensing face a particular gap. A no-action letter bound only the Bureau. A state regulator or attorney general could still act on the same conduct, unless the relief was an approval under one of the three safe harbors. The proposal’s interest in agreements with state authorities was an attempt to close that gap. A nonbank firm would need to track the description it filed with each authority. A change approved by one might breach the description held by another.

Firms outside the program

Firms that never applied still faced a version of the question. If a competitor or a provider held relief, the firm might be offered a product built on it. Its due diligence would need to check what the relief covered, whether the product offered matched the description, and what would happen if the relief ended. A product sold as compliant because of relief held by someone else is only as compliant as that relief’s current scope.

With hindsight

The Bureau finalized the December 2018 proposal in two documents, published September 13, 2019 and applicable from September 10, 2019. It received 31 unique comments on the proposal and 29 covering the sandbox part. Industry commenters supported it. All but one of the consumer group commenters opposed the no-action letter part, and the two groups of state attorneys general split. The final no-action letter policy was characterized solely as a general statement of policy. The no-action letter part became a revised Policy on No-Action Letters. The sandbox part became the Policy on the Compliance Assistance Sandbox. The final sandbox dropped exemptions by order. The Bureau said it intended to propose a legislative rule for exemptions instead.

Then the relief began to end, and it ended through product change. The Bureau granted Payactiv a Sandbox Approval Order in December 2020, a temporary safe harbor under the Truth in Lending Act and Regulation Z for certain earned wage access products. On June 3, 2022, the Bureau told Payactiv it was considering termination over public statements wrongly suggesting a Bureau endorsement. On June 21, 2022, Payactiv said it planned to change its fee model. The approval covered only existing products. Payactiv asked for termination so it could change its fees quickly, and the Bureau terminated the order.

The Bureau’s January 2025 policy statement gave a similar account of Upstart Network’s no-action letter, first granted in 2017 and extended in November 2020. When Upstart wanted to change its model substantially, it applied for a modification. The Bureau did not have enough time to review it, so Upstart asked for termination, and the request was granted.

Those two terminations followed the same sequence. Each holder wanted to change the product, each change fell outside the described aspects, and amending the relief would have taken longer than the holder wanted to wait. In both cases the holder, not the Bureau, asked for the termination. The Bureau’s January 2025 statement later cited both cases as examples of the challenges it had experienced under the earlier policies.

In a statement published September 27, 2022, the Bureau said its Paperwork Reduction Act authorization for both policies would expire and the policies were rescinded as of September 30, 2022. It said they had not advanced their objective and had failed to meet appropriate standards for transparency and stakeholder participation.

On January 10, 2025, the Federal Register published the Bureau’s new no-action letter policy statement, which said a sandbox approval policy was being reissued at the same time. The new conditions barred letters on a topic for a single firm, barred advertising a letter, and put applications on a public docket for 60 days of comment. They also provided that a letter would be automatically rescinded when a product changed beyond its description, unless the holder obtained an amended letter.

That version did not last. On May 12, 2025, the Bureau withdrew both January 2025 policy statements, the no-action letter policy and the sandbox approval policy, in a single Federal Register notice that also withdrew many other guidance documents. As checked on November 14, 2026, a Federal Register search found no later Bureau document reissuing either policy. What held up is the feature this article turns on: relief belonged to a described product, and change ended it.

Spotted an error? Tell us and we will check it and correct the article where needed.

  • CFPB
  • no-action letters
  • regulatory sandbox
  • fintech
  • change management