OMB’s 2019 identity memo made every agency answer for the services it reused
M-19-17 told agencies to reuse other agencies’ credentials, shared sign-in services, approved products and a NIST standard that would keep changing. Reuse saved building effort, but each agency still had to hold evidence that what it relied on met the level it chose.
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
M-19-17 told agencies to reuse other agencies’ credentials, shared sign-in services, approved products and a NIST standard that would keep changing. Reuse saved building effort, but each agency still had to hold evidence that what it relied on met the level it chose.
- The week On May 21, 2019, the Office of Management and Budget issued memorandum M-19-17 to the heads of executive departments and agencies, titled “Enabling Mission Delivery through Improved Identity, Credential, and Access Management.”
- Accepting another agency’s PIV credential The memo told agencies to support cross-government identity federation by identifying and resolving obstacles to accepting PIV identity assertions from other agencies.
- Handing public identity to a shared service The fifth section of M-19-17 dealt with the public.
- Following a standard that would keep moving The phrase “NIST SP 800-63-3 and any successive versions” did a lot of work.
- Buying from lists and approved vehicles The acquisition part of the fourth section told agencies to require HSPD-12 and FIPS 201 compliance in contracts where contractors need access to federally controlled facilities or systems, to confirm that products and services bought for HSPD-12 and ICAM comply with OMB policy, NIST standards and supporting technical specifications, and to use approved Best in Class and Tier 2 contract vehicles, or federally provided shared services, to procure digital certificates.
- Doors, badges and the access nobody reconciles The memo treated physical and logical access as one problem.
- When the agency is the one being reused M-19-17 also ran reuse in the other direction.
- The file a reviewer would ask for Taken together, the four kinds of reuse and the reverse case point to a concrete record.
- The strongest objection: reuse was the right call The objection to our reading is straightforward.
- Who carried the heaviest load The burden of verification under M-19-17 was not the same for every agency.
- What the memo got right about the perimeter It would be unfair to close without crediting the memo for the part of its argument that held up best.
- With hindsight The perimeter argument carried forward.
The week
On May 21, 2019, the Office of Management and Budget issued memorandum M-19-17 to the heads of executive departments and agencies, titled “Enabling Mission Delivery through Improved Identity, Credential, and Access Management.” It set out a new federal identity, credential and access management (ICAM) policy in place of earlier guidance. The memo rescinded five earlier documents: M-04-04 on e-authentication, M-05-05 on electronic signatures, M-06-18 on buying products for Homeland Security Presidential Directive 12 (HSPD-12), M-11-11 on continued HSPD-12 implementation, and a 2011 OMB memo on accepting externally issued identity credentials. National security systems were outside its scope, though OMB encouraged their owners to use it where appropriate.
The memo had six sections. The first defined identity broadly: a person, a device, a non-person entity or an automated technology taking part in a transaction with a federal subject or resource. It separated enterprise identity, meaning the employees, contractors, partners and devices an agency manages, from public identity, meaning the people an agency serves but does not manage.
The second section retired the old “Levels of Assurance” model. Agencies were told they must implement NIST Special Publication 800-63-3 “and any successive versions,” in combination with the other identity publications from NIST, the Office of Personnel Management and the Department of Homeland Security. The third kept HSPD-12 and the PIV card standard, FIPS 201-2, in place but said the government must shift its focus “from managing the lifecycle of credentials to the lifecycle of identities.” Agencies were to require PIV credentials as the primary means of identification and authentication for employees and contractors, use Derived PIV credentials, revoke access and credentials in a timely manner when people leave, and accept other agencies’ PIV credentials rather than issue new ones.
The fourth section was titled “Shifting the Operating Model beyond the Perimeter.” It said agencies must move from managing access inside and outside the perimeter to using identity as the basis for managing the risk of each attempt to reach a federal resource. It required each agency to designate an integrated ICAM office or team, keep a single ICAM policy, process and technology roadmap, and manage the digital identities of devices, robotic process automation tools and artificial intelligence. The fifth section covered services to the public: proofing aligned with NIST, collecting only the personal data needed, and using federal or commercial shared services, to the extent available, to deliver identity assurance and authentication. The sixth assigned tasks to Commerce, the General Services Administration, OPM and DHS, several with deadlines counted from the memo’s issuance (three months, six months or one year).
Those tasks were specific. Commerce was to publish and maintain a roadmap for new and updated NIST identity guidance and to develop, with GSA, criteria for accrediting products and services that meet the SP 800-63 assurance levels. GSA was to publish a roadmap for its identity solutions and shared services, a catalog of existing solutions agencies could use at once, and to keep the FIPS 201 evaluation process and Approved Products List. OPM was to develop a vetting and credentialing model and update eligibility requirements for HSPD-12 credentials. DHS was to align facility security standards with PIV policy and publish a security and privacy control overlay for physical access control systems.
M-19-17 is remembered, when it is remembered, as the memo that moved federal identity policy off the perimeter. That reading is accurate but incomplete. Its instruction with the widest practical reach was to reuse. Agencies were told to accept other agencies’ PIV credentials rather than issue their own, to federate with partners instead of building new accounts, to use shared services to proof and authenticate the public, to buy from approved product lists and government-wide contract vehicles, and to follow a NIST standard “and any successive versions” rather than write their own.
Reuse is sound policy. It cuts cost, reduces the number of identity stores an attacker can target, and spares the public from proving who they are again at every agency. But reuse has a property that the memo did not spell out. When an agency relies on something it did not build, it still owns the risk decision, and it needs evidence that the thing it relies on does what it claims. The memo moved effort from building to verifying. An agency that read it as permission to stop thinking about identity once it had signed an interagency agreement had no defense if a shared service turned out not to meet the standard it advertised, and GSA’s inspector general later found that one had not.
We break the argument into four kinds of reuse that M-19-17 required or encouraged: reusing other agencies’ workforce credentials, reusing shared services for the public, reusing a standard that would keep changing, and reusing commercial products and contract vehicles. Physical access, where several of these meet, gets its own section. Each one moved a piece of work from the agency to someone else, and each one left behind a piece of evidence the agency still had to hold. We then look at the reverse case, where an agency’s own data and credentials are what others reuse, describe the record a reviewer would ask for, take the strongest objection to this reading, and set out how the burden differs by kind of agency, before closing on what the memo got right.
Accepting another agency’s PIV credential
The memo told agencies to support cross-government identity federation by identifying and resolving obstacles to accepting PIV identity assertions from other agencies. It listed what that meant: electronic verification of PIV assertions from other agencies, and accepting existing, valid PIV credentials issued by other agencies, rather than issuing new ones, for both logical and physical access where authorized. It also called for agreements where required to make federation work.
On paper this is a saving. An employee detailed from one department to another, or a contractor supporting several agencies, need not be enrolled, proofed and issued a card twice. The receiving agency skips the enrollment station, the background check paperwork and the card stock. The issuing agency’s proofing and vetting stand in for its own.
What the receiving agency cannot skip is revocation. The memo was explicit that agencies must manage access control, “including the ability to revoke access privileges, when no longer authorized,” and revoke or destroy credentials in a timely manner, so that a separated employee or contractor, or a lost credential, does not leave a door open. When an agency accepts a card it did not issue, two revocation events matter. The issuing agency revokes the certificate when the person leaves that agency. The receiving agency removes the access it granted when the person no longer needs it. The first can happen without the second, and the second can lag the first. A card that is still valid for its issuer may no longer be appropriate for the agency that accepted it, and a card that its issuer has revoked should stop working everywhere.
So the evidence the receiving agency needs is a record of each external credential it accepts, the access it granted against that credential, the check it makes against the issuer’s revocation status at each use, and the event that ends the access on its own side. The issuing agency supplies none of that. It has to be built into the relying systems, and it has to be testable. An auditor may ask for a list of people from other agencies with access to a system and the date each person’s need was last confirmed. If the relying agency can produce that list only by asking each issuer, it has saved the cost of issuing cards and moved the cost into a much harder reconciliation.
The memo’s HSPD-12 section added a further layer. It asked agencies to build capabilities, aligned to the federal ICAM architecture and the Continuous Diagnostics and Mitigation program, that enable continuous vetting and evaluation of the fitness of people subject to HSPD-12, so that access decisions could be risk-adaptive. We would expect continuous vetting to produce its signals at the issuing agency or in the vetting system rather than at the relying agency. A relying agency only benefits from those signals if it consumes them. The design question for a CIO in 2019 was therefore where the revocation and vetting signals would arrive in the relying agency, who would act on them, and how fast. That is an integration and operations question, and the memo assigned the outcome to each agency, not to the issuer.
A useful test, then and now, is to pick a person who left another agency last month and held access to one of the relying agency’s systems through their old card. The relying agency should be able to show when that access stopped and what triggered it. If the honest answer is that the access stopped only because the issuer revoked the certificate, and the relying agency took no action of its own, the relying agency has outsourced a decision the memo left with it. If the answer is that nobody knows, the agency has found the gap before an auditor or an attacker did.
Handing public identity to a shared service
The fifth section of M-19-17 dealt with the public. Agencies were to make sure identity proofing for public-facing digital services aligned with NIST guidance, limit the personal data collected to what is legally authorized, relevant and reasonably necessary, let people bind their own authenticators to their digital identity, and use existing credentials and federations that met the agency’s acceptable risk level instead of standing up new ones. Then came the instruction with the largest operational consequence: agencies “shall use Federally provided or commercially provided shared services, to the extent available, to deliver identity assurance and authentication services to the public,” and those shared services “shall align with NIST SP 800-63 security and privacy requirements.”
The memo also encouraged sharing. With consent and privacy protections, agencies should share proofing confirmations so that people need not submit the same identity data twice, and where consent could not be obtained, agencies should consider changing their Privacy Act system of records notices to add routine uses permitting the disclosure of proofing information. It asked agencies to collect feedback from the public on whether other service providers were in demand, and to prefer shared service providers that use more than one solution, to keep working if one is compromised or fails.
Read together, these instructions split one function into two roles. The shared service does the proofing and authentication. The agency decides what assurance level its service needs, relies on the shared service’s assertion that a person was proofed to that level, and grants access to benefits, records or payments on that basis. The agency’s risk decision is only as good as the assertion it relies on.
That creates a specific evidence need. An agency choosing Identity Assurance Level 2 for a benefits application because of the potential harm of a false claim needs to know that the proofing it receives actually meets IAL2 as NIST defines it. A signed agreement saying so is a statement, not evidence. The evidence is a description of the proofing steps the service performs, mapped to the requirements of SP 800-63A, and some way of confirming that description is true, whether an independent assessment, the provider’s own documented conformance testing, or the agency’s own review. The agency also needs to know what happens when the provider changes its process, and to be told.
The memo’s preference for providers with more than one solution points at a second gap. A shared service is a concentration of risk. When an agency sends its public through one provider, an outage, a fraud campaign or a finding of noncompliance at that provider becomes the agency’s problem at once. The agency’s continuity plan for public sign-in needs to say what it does if the provider is unavailable or found not to meet the level the agency selected. The memo did not require such a plan in those words. It told agencies to prefer resilient providers.
Following a standard that would keep moving
The phrase “NIST SP 800-63-3 and any successive versions” did a lot of work. It tied every agency’s identity program to a document that NIST would revise, without a fresh OMB instruction each time. The memo also told agencies to incorporate Digital Identity Risk Management into existing federal processes as outlined in SP 800-63, select assurance levels commensurate with risk, update legacy e-authentication risk assessments to move away from the old Levels of Assurance model, and send feedback on their implementation to the Federal CIO Council, the Federal Privacy Council and NIST to improve the guidelines.
Pointing at a living standard saves policy effort. It also means an agency’s compliance can lapse without anything changing inside the agency. When NIST publishes a new revision, the agency’s risk assessments, assurance level selections and any reliance on a provider’s conformance claims were written against the old text. The memo set no transition rule. It did not say how long an agency had to move to a successor version, or how an agency should treat a provider that conformed to the previous one.
For a CIO, that turns into a tracking duty. Each system that authenticates people needs a recorded assurance level selection, the version of SP 800-63 it was made under, and the reasoning. When a new version appears, someone has to compare the changes against those records and decide which systems need new assessments. That work falls on the agency even where the system itself is a shared service, because the selection of the assurance level is the agency’s decision. The memo also told Commerce to publish a roadmap with timelines and milestones for new and updated NIST identity guidance within six months of the memo. A roadmap tells agencies when change is coming. It does not do the reassessment for them.
There is a further point about the retirement of Levels of Assurance. The old model gave each transaction a single number that drove the technical requirements. The memo’s footnote said “the concept of LOA as a single ordinal” was retired, and that the updates to SP 800-63-3 “separate the individual elements of identity assurance into discrete, component parts.” GSA’s inspector general later described the two main parts as an Identity Assurance Level for identity proofing and an Authenticator Assurance Level for authentication, each with three levels an agency selects from. That change multiplied the number of decisions an agency had to record for each service: an identity assurance level, an authenticator assurance level and, where federation was used, the federation requirements as well. Each is a choice the agency owns and must be able to explain.
Buying from lists and approved vehicles
The acquisition part of the fourth section told agencies to require HSPD-12 and FIPS 201 compliance in contracts where contractors need access to federally controlled facilities or systems, to confirm that products and services bought for HSPD-12 and ICAM comply with OMB policy, NIST standards and supporting technical specifications, and to use approved Best in Class and Tier 2 contract vehicles, or federally provided shared services, to procure digital certificates. It also pointed agencies to the Continuous Diagnostics and Mitigation program to speed up the purchase and deployment of ICAM tools in its second phase or later phases.
The architecture part asked agencies to rationalize their ICAM capabilities, deciding which to keep, replace, retire or consolidate, and to make sure deployed capabilities are interchangeable, use commercially available products, and use open application programming interfaces and commercial standards to promote interoperability. GSA, for its part, was to maintain the FIPS 201 evaluation process and its Approved Products List, and to publish within three months a consolidated catalog of ICAM solutions and shared services agencies could use immediately.
A listed product or an approved vehicle lowers the cost of a purchase decision. It does not settle whether the product, as configured in a particular agency, does what the agency’s risk decision requires. An approved card reader still has to be connected to a physical access control system that checks revocation. A certificate bought through a government-wide vehicle still has to be issued only to the right people and revoked when they leave. An identity platform bought through the CDM program still has to be connected to the agency’s authoritative sources of personnel data.
The memo also asked agencies to manage the identities of devices, non-person entities and automated technologies, including robotic process automation and artificial intelligence, so that each digital identity is “distinguishable, auditable, and consistently managed across the agency,” with mechanisms to bind, update, revoke and destroy their credentials. We think that requirement could not be met by buying from a list. It needed an inventory of service accounts and automated processes, an owner for each, and a way to retire them. For an agency that had automated back-office tasks with scripts running under shared or personal accounts, this was new work with no product shortcut.
Doors, badges and the access nobody reconciles
The memo treated physical and logical access as one problem. Agencies were to require PIV as the primary means of authentication to “Federally controlled facilities and secured areas” as well as to information systems, and to implement PIV for physical access in line with the Interagency Security Committee’s risk management standard for federal facilities and NIST SP 800-116 Revision 1 on PIV in facility access. Its footnote made a point that is easy to pass over: physical access control systems are information systems, including servers, databases, workstations and network appliances. DHS was tasked with publishing a security and privacy control overlay for those systems, developed in consultation with GSA, OPM, OMB and Commerce.
Physical access is where the reuse of other agencies’ credentials becomes visible. A visitor from another department presents a PIV card at a turnstile. The reader is on GSA’s Approved Products List. The card is valid for its issuer. Whether the person should pass depends on whether the receiving agency registered that card, for that person, with an end date, and whether its access control system checks the card’s revocation status at each use or only at registration.
These systems may be run by facilities or security offices rather than by the CIO, and in a building shared by several tenants they may be run by one agency for all of them. The memo’s governance section named the senior officials responsible for physical security as members of the agency’s integrated ICAM structure. That inclusion mattered. Without it, the identity team could build careful revocation for applications while the door readers kept honoring cards for people who had left.
The evidence needed is the same as for applications. Each physical access system should be able to list the credentials registered to it, the agency that issued each, the access granted, the check against revocation, and the date access ends. It should be inside the agency’s security authorization process, with an owner who can answer for its configuration. Where the system is run for several agencies, each tenant should know which of its people are registered and be able to remove them. We think the DHS overlay the memo called for was meant to give these systems the same control baseline as any other system holding personal data, and an agency that treated its badge system as building equipment rather than as an information system would have missed that intent.
Physical access also showed how the memo’s two lifecycles, of credentials and of identities, can come apart. A PIV card has a certificate lifecycle managed by its issuer. A person’s need to enter a building has a lifecycle managed by a supervisor, a contract or an assignment. The memo’s call to shift attention from credentials to identities was, in a building, a call to connect the badge system to the records that say why a person is there and until when.
When the agency is the one being reused
M-19-17 also ran reuse in the other direction. Its architecture section said that agencies that are authoritative sources for attributes used in identity proofing, giving the Social Security number as an example, “shall establish privacy-enhanced data validation APIs for public and private sector identity proofing services to consume,” where selected by OMB and permissible by law, to improve the assurance of identity verification “based on consumer consent.” Those agencies, with OMB, were to set standard processes and terms of use for the services that wanted to call the APIs.
For an agency in that position, the memo created a new kind of product with outside customers. PIV assertions were consumed by other federal agencies. A validation interface consumed by private identity proofing companies is different. Its outputs feed decisions the source agency never sees, made by firms it does not supervise, about people who may never have dealt with the agency directly.
The evidence this creates belongs to the source agency. It needs terms of use that say who may call the interface and for what purpose, a way to confirm that each call carries the individual’s consent, records of who asked about whom, limits on what is returned, and a way to cut off a consumer that breaks the terms. It needs monitoring that can spot a caller using the interface to test stolen identities at scale. It also needs to explain, in its Privacy Act notices, the routine uses that permit the disclosures. Each of these is a control the source agency must run and be able to show. None of them is satisfied by the consuming service’s own assurances.
The same applies, on a smaller scale, to any agency whose PIV assertions other agencies accept. The issuing agency’s revocation speed becomes part of every relying agency’s security. If an issuer is slow to revoke a certificate after a separation, the gap shows up in systems the issuer does not run. An issuing agency that wants others to rely on its credentials should be able to state how quickly it revokes after a separation and to show its record.
This is the reverse side of the argument in the earlier sections. Reuse shifts effort from the party that would have built something to the party that now relies on it. When an agency is the thing being relied on, it carries the duty to make its own behavior measurable, because the parties relying on it have only its word.
The file a reviewer would ask for
Taken together, the four kinds of reuse and the reverse case point to a concrete record. For each system that authenticates people or processes, an agency acting on M-19-17 would want a file that a reviewer could read without interviewing anyone.
The file would state the system’s identity and authenticator assurance levels and, where federation is used, the federation requirements chosen, the version of SP 800-63 under which they were selected, the potential harms considered, and the official who accepted the residual risk. It would list every outside party the system relies on for identity: other agencies whose credentials it accepts, shared services that proof or authenticate its users, federations it trusts. For each, it would say what claim the system relies on and what evidence supports the claim, with a date.
It would describe how the system learns of revocations and changes from those parties and how fast it acts. It would list the non-person identities that touch the system, such as service accounts, scheduled jobs, automation tools and integration accounts, with an owner and an end date for each. It would record the products and contract vehicles used, and any configuration that departs from the product’s evaluated state.
For public-facing systems, it would show the personal data collected at proofing, the legal authority for each element, any sharing of proofing confirmations and the consent or routine use relied on, and the continuity arrangement if the identity provider is unavailable or found deficient. For systems that expose identity data to others, it would show the terms of use, the consent check and the monitoring.
Much of this information probably existed somewhere in a federal agency in 2019, spread across authorization packages, interagency agreements, contract files and privacy documents. The memo’s requirement for a single agency-wide ICAM governance structure and a single roadmap was an invitation to pull it together. An agency that did so had, in one place, the evidence that reuse required. An agency that did not would find, when a provider’s claims were questioned, that the answer was scattered across offices that had not compared notes.
The strongest objection: reuse was the right call
The objection to our reading is straightforward. M-19-17 was right to push agencies toward reuse. A government in which each agency proofs the public on its own, issues its own workforce credentials and buys its own identity stack would be more expensive, less consistent and harder to secure. The memo’s own text names the risk management duties that remain with agencies. If some agencies treated a signed agreement as the end of their responsibility, the fault lies with them, not with the policy.
We accept this objection’s premises but not its conclusion. The policy was right. Its weakness was that it described the duty to verify in general terms while describing the duty to reuse in specific ones. Agencies “shall use” shared services “to the extent available.” Agencies “shall accept assertions by partners.” The corresponding verification duties appear as broader risk management statements: incorporate Digital Identity Risk Management, make risk-informed decisions, confirm products comply with standards. A program office reading the memo for its action items would find the reuse instructions easier to act on and easier to show compliance with.
A second objection is that the memo could not have specified verification methods without becoming a technical standard itself, and that NIST, not OMB, writes those. That is fair. But the memo did assign roles. It asked Commerce and GSA to develop criteria for accrediting products and services that meet the SP 800-63 assurance levels, and GSA to determine the feasibility of an accreditation capability for ICAM products and services on its vehicles. Those tasks show that the authors saw the need for third-party confirmation. Until such confirmation existed, the agencies relying on shared services were the only ones who could ask for evidence, and many had no process for asking.
A third objection concerns timing. In May 2019, some of the shared services agencies were told to use may themselves still have been developing. Holding a relying agency to a standard of evidence that its provider could not yet supply would have slowed adoption. Again, fair. But that argues for recording the gap, not ignoring it. An agency that wrote down, in 2019, that it was relying on a provider’s statement of IAL2 conformance without independent confirmation, and set a date to revisit it, had a defensible file when questions came.
Who carried the heaviest load
The burden of verification under M-19-17 was not the same for every agency. It depended on what the agency did with the identities it relied on.
Benefit-paying and records agencies. Agencies that pay money or release sensitive records to the public on the strength of an online identity carried the largest exposure. For them, the assurance level selection decided how much fraud and how many wrongful disclosures they would absorb. They needed the clearest evidence that a shared service’s proofing met the level chosen, and the clearest plan for what to do if it did not.
Large departments with many components. The memo asked each agency to designate an integrated agency-wide ICAM office and to keep a single comprehensive roadmap, with bureaus and components adhering to those requirements. For a department with many components, each with its own systems and history, the hard part was getting a single picture of which systems accepted which credentials, at which assurance level, under which version of the standard. The memo named the department’s Chief Operating Officer, or equivalent, as responsible for regular coordination among leaders and mission owners on ICAM.
Small agencies. Small agencies were likely to gain the most from shared services, since few could build their own. They also tended to have the least capacity to examine what they were buying. For a small agency, verification would have to come from somewhere else: a government-wide assessment, GSA’s accreditation work, or a peer agency’s review. Until that existed, a small agency could still record which provider claims it relied on and ask its provider, in writing, how each claim was confirmed.
Agencies working with state, local and tribal governments. The memo told agencies to coordinate with state, local and tribal governments, other entities and individuals to provide identity verification and access control appropriate to the risk of the business function where information sharing or collection is required. Where a federal agency accepts assertions from a state system, or a state accepts federal ones, the same evidence question runs in both directions: what proofing stands behind the assertion, and who will say when it changes.
Contractors. For companies working under federal contracts, M-19-17 reached them through contract clauses requiring HSPD-12 and FIPS 201 compliance for personnel who need routine access. Their own enterprise identity systems were outside the memo, but the PIV cards they carried and the federated access they used were inside it. A contractor supporting several agencies became a test case for whether cross-agency acceptance and revocation actually worked.
Agencies with heavy automation. Agencies that had automated back-office work, such as reconciliations, report generation and data transfers between systems, faced the memo’s requirement on non-person identities in its fullest form. Each automated process that signs in to a system is an identity the memo said must be distinguishable, auditable and consistently managed. An agency with a large number of scheduled jobs and integration accounts had to inventory them before it could manage them, and inventory is slow work that no shared service performs on an agency’s behalf. The same agencies carried the greater exposure if an automation account with broad access was forgotten after the person who built it moved on.
What the memo got right about the perimeter
It would be unfair to close without crediting the memo for the part of its argument that held up best. The fourth section’s claim that hardening the perimeter is important but insufficient, and that identity must be the underpinning for managing the risk of each access attempt, was carried into federal zero trust policy: OMB’s M-22-09 of January 2022 cites M-19-17 by name. The memo’s requirement to manage the identities of devices and automated technologies addressed software that acts on its own, such as robotic process automation and artificial intelligence, which it named in 2019. Its insistence on a single governance structure, with the CFO, human resources, general counsel and acquisition at the table alongside the CIO and CISO, recognized that identity runs through hiring, separation, contracting and payment as well as through IT.
What it underplayed was the evidence that reuse demands. The memo’s instructions to reuse were concrete, dated and easy to report against. Its instructions to verify were framed as risk management principles, which are harder to measure and easier to defer. Each of the four kinds of reuse in the memo shifted work from building to verifying. For a federal executive reading it in 2019, the useful question was not whether the agency had adopted shared services and accepted federal credentials, but whether it could show, for each system, which outside party it relied on, for what claim, and how it knew the claim was true.
With hindsight
The perimeter argument carried forward. On January 26, 2022, OMB issued M-22-09, “Moving the U.S. Government Toward Zero Trust Cybersecurity Principles,” which set specific goals for agencies to meet by the end of fiscal year 2024. Its identity pillar required centralized identity management systems that can be integrated into applications, multi-factor authentication enforced at the application layer, phishing-resistant authentication for agency staff, contractors and partners, a phishing-resistant option for public users, and password policies that do not require special characters or regular rotation. M-22-09 cited M-19-17 directly. It noted that M-19-17 requires PIV credentials as the “primary” means of authentication to federal systems, said PIV would not be practical in every case, and permitted phishing-resistant authenticators that did not yet support PIV, such as those based on FIDO2 and Web Authentication.
The standards M-19-17 pointed to moved as its wording anticipated. NIST published FIPS 201-3 in January 2022, superseding FIPS 201-2. It published SP 800-63-4 in July 2025, superseding SP 800-63-3. Because M-19-17 bound agencies to SP 800-63-3 “and any successive versions,” the target moved for every agency without a new memo. When we checked on November 14, 2026, NIST’s publication page for SP 800-63-4 carried a planning note dated 10/06/2026 saying the Implementation Resource Hub for SP 800-63-4 “is now available.”
The clearest test of the reuse argument came from GSA’s own inspector general. In a report dated March 7, 2023 (JE23-003), the GSA Office of Inspector General found that GSA had misled customer agencies by failing to communicate Login.gov’s known noncompliance with SP 800-63-3. Despite GSA officials’ assertions that Login.gov met Identity Assurance Level 2, the report said Login.gov had never included a physical or biometric comparison for its customer agencies. It found that GSA knowingly billed IAL2 customer agencies over $10 million for services that included alleged IAL2 services not meeting the standard, and that GSA notified customer agencies on February 3, 2022, that its services were not compliant. The report quoted M-19-17’s “must implement” instruction on SP 800-63-3 and said any claim that GSA had flexibility in meeting the NIST standard “lost viability in May 2019” when OMB issued it. It also quoted a GSA funding proposal that described Login.gov as providing shared services “in accordance with M-19-17.” GSA management agreed with the findings and the five recommendations.
That finding is the risk this article describes, realized. The agencies relying on Login.gov had followed M-19-17’s instruction to use a federally provided shared service. The assurance they relied on was not there. The report placed responsibility on GSA, which made the claims and lacked controls over the program. But under M-19-17 the assurance level selections were the customer agencies’ own, and the report shows them relying on GSA’s statements that those levels were met.
What held up, then, was the memo’s architecture: identity as the basis for access decisions, reuse over duplication, and binding agencies to NIST’s current guidance. What did not hold up was the assumption, implicit in its structure, that the party providing a shared identity service would be a reliable source of evidence about its own conformance. As checked on November 14, 2026, SP 800-63-4 is the current version of the guidelines M-19-17 binds agencies to, and the lesson of the Login.gov finding applies to any provider they rely on under it: an assertion about assurance is a claim to be checked.
Spotted an error? Tell us and we will check it and correct the article where needed.