NIST’s August 2017 SP 800-53 draft made controls more outcome-based and dropped “Federal” from its title

NIST opened SP 800-53 Revision 5 for public comment on August 15, 2017. The draft merged security and privacy controls, phrased controls as outcomes and dropped “Federal” from the title. In our reading, it moved the hard choices to whoever selects and assigns controls.

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

NIST opened SP 800-53 Revision 5 for public comment on August 15, 2017. The draft merged security and privacy controls, phrased controls as outcomes and dropped “Federal” from the title. In our reading, it moved the hard choices to whoever selects and assigns controls.

  1. The week In August 2017, the National Institute of Standards and Technology released the initial public draft of Special Publication 800-53, Revision 5, for comment.
  2. The audience: a catalog written for anyone who builds or runs a system The new title dropped “Federal.”
  3. The control statement: outcomes without a named actor The second change was in how each control was written.
  4. The catalog’s contents: privacy controls folded in, with their own owner Revision 4 kept privacy controls in a separate appendix, Appendix J, as NIST’s Revision 5 page still notes in offering a mapping from it [6].
  5. The selection layer: baselines kept, tailoring guidance moved out The fourth change was the separation of control selection from the controls.
  6. How the draft reached different organizations The draft was not binding.
  7. What the draft left open The draft left several questions unsettled.
  8. Objections, in order Legally, that is right.
  9. What an agency CISO or system owner could check The draft’s changes are now settled in the final catalog.
  10. With hindsight The schedule in the draft’s notes did not hold.

The week

In August 2017, the National Institute of Standards and Technology released the initial public draft of Special Publication 800-53, Revision 5, for comment. The draft gave its public comment period as August 15 through September 12, 2017 [1]. NIST’s publication page lists comments as due on September 12, 2017 [2]. The document ran to 494 pages and carried the title Security and Privacy Controls for Information Systems and Organizations.

The title was itself a change. Revision 4, published in April 2013 and updated in January 2015, was titled Security and Privacy Controls for Federal Information Systems and Organizations [3]. The new version removed the word “Federal.” Revision 4 was the catalog federal agencies used to select controls for their systems under the Federal Information Security Modernization Act.

The publication was produced by the Joint Task Force. Its acknowledgments thanked an interagency technical working group and named the Departments of Commerce and Defense, the Office of the Director of National Intelligence and the Committee on National Security Systems. They also thanked members of a subcommittee of the Federal Privacy Council [1].

The notes to reviewers explained its purpose. They cited a 2017 Defense Science Board task force report on cyber defense and its assessment of vulnerabilities in U.S. critical infrastructure. They said the revision aimed to provide safeguards for all types of computing platforms. The list included general purpose computing systems, cyber-physical systems, cloud and mobile systems, industrial and process control systems, and Internet of Things devices [1].

The notes listed six major changes. It made the controls more outcome-based by changing their structure. It merged the privacy controls into the security control catalog. It separated the control selection process from the controls themselves. It promoted integration with other risk management and cybersecurity approaches, including the Cybersecurity Framework. It clarified how the two disciplines relate. And it added new controls based on threat intelligence and attack data [1].

The notes also described a cost of the separation. A significant amount of tailoring guidance and other informative material had been eliminated from the publication. NIST said that content would move to other publications, such as SP 800-37 on the Risk Management Framework, in their next update cycle. Until then, it would remain available in Revision 4 [1].

The document organized its controls into twenty families. Two of them, Individual Participation and Privacy Authorization, were privacy families. A footnote said seventeen of the twenty families aligned with the security requirements in FIPS Publication 200, and three covered other matters, including program management [1].

The statement of applicability did not match the new title. Section 1.1 said the publication established controls for federal information systems and organizations, and that their use was mandatory under FISMA. It added that the controls could be implemented within any organization or system that processes, stores or transmits information. It encouraged state, local and tribal governments and private sector organizations to consider using them [1].

The notes set out a schedule. NIST anticipated producing a final draft in October 2017 and publishing the final version no later than December 29, 2017 [1]. The publication page later recorded that the comment period had closed. It added a planning note on February 9, 2018 pointing readers to the current publication schedule proposed by NIST, which it said may be subject to change [2].

A control catalog is easy to read as a list. The August 2017 draft of SP 800-53 Revision 5 changed less in the list than in the frame around it. It changed who the catalog spoke to, how each control was phrased, what the catalog held and where the decision about which controls to use would be made.

Each of those changes moved a decision. Under Revision 4, the catalog spoke to federal agencies, phrased controls as duties of the organization or the system, kept privacy in an appendix and bundled selection guidance with the controls. The draft loosened each of those links. In our reading, it turned the catalog into a common vocabulary and left the binding choices to whoever selected and assigned the controls.

Our argument is that the draft moved the hard choices out of the control text and into the selection and assignment layer around it. We take the draft in four parts:

  1. The audience: a catalog written for anyone who builds or runs a system.
  2. The control statement: outcomes without a named actor.
  3. The catalog’s contents: privacy controls folded in, with their own owner.
  4. The selection layer: baselines kept, tailoring guidance moved out.

The first part concerns who could use the catalog. The second concerns how a control reads. The third concerns what the catalog contained. The fourth concerns how anyone decided which controls applied. Together they cover the draft from its intended reader to the decision that reader would make. After the four parts we look at how the draft reached different organizations, at what it left open, at objections, and at what an agency CISO or system owner could check.

A note on sources and our reading. We describe the draft from its own text and NIST’s publication pages. We describe later versions from NIST’s final publications and their pages. The draft was a draft, and its contents changed before the final version. Statements about what a version required of whom are our reading. Agencies and contractors should rely on the current NIST text and their own counsel.

The audience: a catalog written for anyone who builds or runs a system

The new title dropped “Federal.” The notes to reviewers said the revision would make safeguards available to a broad base of public and private sector organizations [1]. They named the types of platforms in scope, from general purpose computers to industrial control systems and Internet of Things devices. Revision 4’s title had addressed federal information systems [3].

Applicability stayed federal

The draft’s applicability section kept the federal anchor. It said the publication established controls for federal information systems and organizations and that FISMA made their use mandatory. A footnote defined a federal information system to include systems used or operated by a contractor of an agency, or by another organization on an agency’s behalf [1].

So the legal reach did not change. Agencies were still bound, through FISMA and OMB policy. Contractors operating systems for agencies were still reached through their contracts. What changed was the stated audience. The draft said the controls could be implemented within any organization or system that handles information. It encouraged state, local and tribal governments and private organizations to consider them.

A wider list of readers

The draft’s target audience section named five groups. They included people with oversight duties, such as authorizing officials and chief information officers. They included people who build systems, such as systems engineers, privacy engineers, software developers, integrators and procurement officials. The list also named commercial entities producing components, products, systems and services [1].

Naming builders and vendors as readers matters. Under a catalog addressed to agencies, a vendor learns about controls second-hand, through a contract clause or an assessment. Under a catalog addressed to builders, the vendor can design to the controls directly. The draft’s section on developing systems and components made that point explicitly for the System and Services Acquisition family.

Controls aimed at developers

The draft said a set of controls in the System and Services Acquisition family was directed at developers of systems, components and services. That applied whether development happened inside the organization or through industry partners such as manufacturers, vendors and integrators. It named SA-8, SA-10, SA-11, SA-15, SA-16, SA-17, SA-20 and SA-21 as the affected controls [1].

In our reading, this was the architectural core of the audience change. An agency cannot build a trustworthy system from untrustworthy components. The draft gave agencies language to state requirements to the people building those components. It also gave the builders a public reference for what would be asked of them.

Cyber-physical and connected devices

The platform list in the notes included cyber-physical systems, industrial and process control systems and Internet of Things devices [1]. Those systems differ from office IT. They may run for decades, lack the processing power for some controls and serve safety functions that patching can disrupt. A catalog written for them has to allow tailoring that would look odd on a server.

For a public agency, those systems include building controls, water and transit equipment and field sensors. For a health system, they include connected clinical devices. Naming them in scope did not make the controls fit them automatically. It did tell owners of such systems that the catalog intended to cover them, and that tailoring was expected.

What a wider audience costs

A catalog written for everyone must avoid assumptions that hold only for agencies. That pushes text toward the general. The draft’s notes admitted as much when they said tailoring guidance had been removed. A non-federal reader gains access but loses the federal context that told an agency how to apply each control. That reader has to supply the context from its own rules and risks.

For a state agency or a hospital, that is a real cost. The draft itself ran to 494 pages. Without a baseline that fits, an organization must decide which controls matter. The draft’s Appendix F said nonfederal organizations may use overlays to tailor selection to the laws that apply to them [1]. Building an overlay is work.

Why the audience shift mattered to agencies too

Agencies gained from a wider audience. A shared catalog lets an agency’s requirements match the vocabulary its vendors already use. It lets a state program that receives federal data map its own controls more easily. The draft’s section on common foundations said a unified framework provided a basis for reciprocal acceptance of authorization decisions [1]. Shared language makes reciprocity cheaper.

The control statement: outcomes without a named actor

The second change was in how each control was written. The notes said the draft made controls more outcome-based by changing their structure [1]. The final version, published in 2020, described the same change more precisely. It said controls became more outcome-based by removing the entity responsible for satisfying the control, the information system or the organization, from the control statement [5].

What the structure looks like

The draft described five sections for each control. A base control prescribes a capability. Supplemental guidance adds non-prescriptive information. Control enhancements add function or strength. Related controls point to others that affect implementation. References list applicable standards [1].

The draft’s worked illustration was AU-4, Audit Storage Capacity. The control statement began with a verb: allocate audit record storage capacity to accommodate an organization-defined retention requirement. Nothing in the sentence said who allocates. The draft said organizations designate responsibility for control design, implementation, management and monitoring [1].

Assignment becomes a separate decision

Removing the actor makes allocation explicit. For AU-4, the storage might be allocated by the agency’s own operations team, by a cloud provider, or by a managed service. Under the new phrasing, the control does not presume an answer. The system security plan must give one.

In our reading, this change had large operational effects. The draft itself listed cloud and mobile systems among the platforms in scope, and a footnote counted contractor-operated systems as federal systems. A control phrased as a duty of “the organization” could hide the fact that a provider performed it. A control with no named actor forces the plan to say who does it and how the agency knows.

Common and inherited controls

The draft’s tailoring appendix listed identifying and designating common controls as a tailoring activity [1]. A common control is implemented once and inherited by multiple systems. An agency-wide identity service, for example, can satisfy authentication controls for dozens of applications. Outcome phrasing makes inheritance easier to express, because the control text does not tie the outcome to one system.

Inheritance has a weakness. A system owner who inherits a control depends on another team’s evidence. If the identity service changes, every inheriting system is affected. The plan must record the dependency, and the assessment must reach the provider of the common control. The catalog cannot do that work. The selection and assignment records must.

Parameters as the binding detail

The draft kept organization-defined parameters. Assignment statements let an organization supply values, such as a retention period. Selection statements limit the choice to a list. The draft said the values, once specified, become a permanent part of the control, and the implementation is assessed against the completed statement [1].

That means the binding text of a control is partly written by the organization. Two agencies can both implement AU-4 and mean different retention periods. Two vendors can both claim to support a control and assume different parameters. In our reading, outcome phrasing increased the weight carried by parameters, because the control text itself said less about context.

Evidence for an outcome

Outcome phrasing changes what evidence looks like. A control that names the system as the actor suggests a technical setting as evidence. A control that names the organization suggests a policy or a record. A control with no actor accepts either, depending on how it was implemented. The assessor must first learn how the control was implemented before choosing what to test.

That puts weight on the implementation description in the plan. If the plan says storage capacity is set by a cloud provider’s service, the evidence is the provider’s configuration and its attestation. If the plan says the agency’s operations team monitors capacity, the evidence is monitoring records. The control text is the same. The test differs.

What outcome phrasing does not change

Outcome phrasing does not reduce the work. A control that says “allocate storage” still has to be done, documented and assessed. What it changes is where ambiguity sits. Under the older phrasing, ambiguity could hide in who “the organization” meant. Under the new phrasing, the plan must resolve it explicitly or the assessment will find the gap.

A test for any control

One way to read the change is as a test. For each selected control, an organization should be able to name the actor, the parameter values, the evidence and the assessor. If any of those is missing, the control is not yet specified. The draft’s structure made the gaps visible. It did not fill them.

The catalog’s contents: privacy controls folded in, with their own owner

Revision 4 kept privacy controls in a separate appendix, Appendix J, as NIST’s Revision 5 page still notes in offering a mapping from it [6]. The draft folded them into the main catalog. The notes said this created a consolidated and unified set of controls, while providing summary and mapping tables for privacy-related controls [1]. Two of the twenty families, Individual Participation and Privacy Authorization, were privacy families.

Privacy and joint designations

Appendix F gave a consolidated view of privacy controls. It marked each relevant control or enhancement as privacy or joint. For controls marked privacy, privacy programs had sole authority to select and oversee them. For joint controls, authority and oversight were shared with information security programs [1].

The draft gave training as an example. For AT-2, a privacy program could run separate privacy training or coordinate with the security program on a combined course [1]. The point of the designation was ownership. It told an agency which program decided and which program answered for the result.

A footnote said the source of this authority was OMB Circular A-130 [1]. In our reading, the draft built the federal division of labor between security and privacy programs into the catalog’s structure. An agency that adopted the integrated catalog also adopted a map of who owned each control.

Required, situational and discretionary

Appendix F also tagged privacy controls by selection criteria. Required controls must be selected based on legal, regulatory or policy requirements. Situationally required controls apply only in specific circumstances. Discretionary controls may be selected based on privacy risk assessments [1].

These tags did a job baselines did for security. A footnote to the baseline appendix said privacy-related controls were not allocated to the low, moderate and high baselines [1]. Privacy selection followed its own logic, driven by law and by the processing of personal information, not by system impact level.

Mapping is not compliance

Appendix F also mapped OMB Circular A-130’s privacy requirements to catalog controls. The draft warned that agencies should not assume implementing the controls meant they had met their A-130 obligations. Agencies might need additional, separate steps [1]. It said the mapping also supported nonfederal organizations that had to meet those requirements under federal contracts or agreements.

That warning applies well beyond privacy. A control catalog describes capabilities. Legal obligations are stated in law and policy. A mapping connects the two but does not merge them. An organization that reports compliance with a law by pointing to a control list has skipped a step the draft told it not to skip.

What integration asked of programs

Integration asked security and privacy programs to share a document and divide its contents. In an agency where the two programs worked separately, that meant agreeing on joint controls, assessment frequency and responses to deficiencies. The draft said organizations should clearly define and delineate responsibility for implementation, assessment and monitoring between the two programs [1].

For a health system or bank that used 800-53 voluntarily, the same structure offered a model. Such an organization has privacy obligations under its own laws. The joint designation gave it a way to record which controls its privacy office owned and which it shared with security.

Assessment and deficiencies across two programs

Shared ownership has a practical edge at assessment time. The draft said responsibility for assessment and monitoring should be clearly delineated between the programs, including the frequency of control assessments and the response to deficiencies [1]. A joint control assessed by the security program may produce a finding the privacy program must fix, or the reverse. Without an agreed route, findings stall between offices.

Personal information in system design

Integration also changed when privacy enters a project. With privacy controls in an appendix, a system team could treat them as a later review. With privacy controls in the main catalog and owned by a privacy program, they enter selection with everything else. In our reading, that pushed privacy questions into design, where data flows and retention are cheapest to change.

A system that collects personal information must answer questions about notice, consent, use limits and individual access. Those answers shape data models and interfaces. Changing them after build costs more than deciding them before. The draft’s structure did not require early decisions. It made them easier to schedule, because the controls arrived in the same list.

New controls beside the privacy ones

The notes said the draft also added new state-of-the-practice controls based on threat intelligence and empirical attack data. They included controls to strengthen cybersecurity and privacy governance and accountability [1]. The developer-facing acquisition controls discussed in the first part belong here too. The catalog grew in both directions, toward privacy and toward the supply side.

The selection layer: baselines kept, tailoring guidance moved out

The fourth change was the separation of control selection from the controls. The notes said this would let the controls be used by different communities, including systems engineers, software developers, enterprise architects and mission or business owners [1]. The draft did not complete the separation. It took a first step.

What stayed

The low, moderate and high security baselines stayed in Appendix D. A footnote said the high baseline was a superset of the moderate, and the moderate of the low. Another footnote said federal organizations tailor baselines in accordance with OMB policy, and that OMB may prohibit tailoring out specific controls [1].

A tailoring appendix also stayed, in shorter form. It listed activities such as designating common controls, applying scoping considerations, selecting compensating controls, assigning parameter values, supplementing baselines and specifying implementation. It said tailoring decisions should be defensible and documented in security and privacy plans [1].

Rationale as the record

The tailoring appendix said organizations should not arbitrarily remove controls from baselines. Decisions should be defensible, based on mission and business needs, a sound rationale and explicit risk-based determinations. It said tailoring decisions and their justification are documented in the security and privacy plans [1].

That makes the rationale a record an assessor will read. A plan that drops a control with the words “not applicable” and no reason gives the assessor nothing to evaluate. A plan that explains why the system does not need the control, or what compensates for it, can be tested. As selection moved toward design, the quality of those explanations became a measure of the selection itself.

What moved

The notes said a significant amount of tailoring guidance and informative material had been eliminated. It would move to SP 800-37 and other publications during their next update cycle, and remain active in Revision 4 until then [1]. For a period, an agency reading the draft would need Revision 4 to get the full tailoring context.

Selection without baselines

The tailoring appendix contained a telling footnote. It said some organizations may use the controls without baselines, for example when applying them in a life cycle-based systems engineering process. In that case controls are selected through a requirements engineering process meant to provide a system capability [1].

In our reading, that footnote described where the draft was heading. Controls become a library. Selection becomes a design activity. An engineer building a system picks controls to meet requirements, much as the engineer picks components. The baseline becomes one input to that choice, rather than the starting point.

Overlays as the middle layer

The draft pointed to overlays as a way to tailor controls for specific missions, technologies, environments or communities of interest [1]. An overlay is a reusable set of tailoring decisions. A community that shares a type of system, such as a class of medical devices or a kind of industrial controller, can agree an overlay once and apply it repeatedly.

Overlays sit between the catalog and a single system’s plan. As selection moved out of the catalog, that middle layer became more important. In our reading, the organizations best placed to benefit from the draft were those that could build or adopt overlays for their common system types. Those that tailored every system from scratch faced heavier work.

Who owns selection

The separation raises a governance question. If selection is a design activity, the people making design decisions are making security and privacy decisions. In an agency, that includes program offices and contractors. The authorizing official still accepts the result, but the content of the selection may be shaped earlier, in requirements documents and contracts.

The draft’s footnote on authorizing officials said that by accepting the security and privacy plans, they agree to the set of controls proposed to meet the security and privacy requirements [1]. Acceptance comes at the end. In our reading, the separation made it more important that security and privacy staff take part at the start, when requirements are written.

Selection for non-federal adopters

A non-federal organization has no OMB policy telling it how to tailor. For it, separation means the catalog brings no default baseline with legal weight. It may adopt NIST’s baselines as a convenience. It may build its own from sector rules. Either way, it owns the choice. The draft gave it a library and left it to supply the selection logic.

The cost of moving guidance

Moving guidance into other documents spreads it across publications on different update cycles. An agency must keep several documents in step. A vendor reading one document may miss guidance in another. That cost is the price of letting different communities use the controls without carrying federal selection rules they do not need.

How the draft reached different organizations

The draft was not binding. It still reached organizations in different ways, depending on how they used the catalog and how soon they planned to adopt the final version.

Agency CIOs and CISOs

For agency security leaders, the draft signaled future work. Control families would change, privacy controls would move into the catalog and tailoring guidance would move elsewhere. Agencies could not adopt a draft as policy. They could plan for the transition, review their control inheritance records and talk to their privacy programs about ownership.

Senior agency officials for privacy

Privacy officials gained a defined place in the catalog. The privacy designation in Appendix F gave them sole authority over certain controls, with OMB Circular A-130 as the source. That made privacy staffing and assessment capacity a direct question. A privacy program that owned controls had to assess them.

Program offices and systems engineers

The draft named systems engineers and developers as readers. For program offices acquiring new systems, it offered language for requirements documents. In our reading, the draft made it easier to write control requirements early, in solicitations and design reviews, instead of retrofitting them before authorization.

Contractors and product vendors

Vendors gained a public reference addressed partly to them. A vendor selling to agencies could design products to support named controls and parameters. It could also expect more developer-facing requirements in contracts, given the draft’s emphasis on the acquisition family.

State and local governments

The draft encouraged state, local and tribal governments to consider the controls. For a state agency that receives federal data, the catalog could already be a reference through its agreements. The title change made that reference read less like a federal-only document. It did not change any state’s obligations.

Banks and health systems

Financial and health organizations answer to their own regulators and frameworks. One that uses 800-53 as a detailed control library behind a higher-level framework faces a specific choice. For such an organization, the draft offered a broader catalog with integrated privacy. It also meant a future remapping exercise when the final version arrived.

Cloud and shared service providers

Providers that host systems for multiple agencies already worked with inheritance. Their customers can inherit physical, environmental and infrastructure controls from them. Outcome phrasing suited that model, because a control without a named actor can be split between provider and customer. It also meant the split had to be written down, control by control, for each customer system.

Auditors and inspectors general

The draft named auditors and inspectors general among its audience. For them, outcome phrasing meant that the assessment would test a completed control statement, including the organization’s parameters and assignments. The quality of those records would decide how assessable each control was.

What the draft left open

The draft left several questions unsettled. It kept baselines inside the publication while saying selection should be separated. It removed tailoring guidance before its new home existed. Its title addressed everyone while its applicability section addressed federal systems. Each of those tensions would need resolving in a later draft.

It also set a schedule it did not meet. The notes anticipated a final version by December 29, 2017 [1]. By February 2018 the publication page carried a planning note pointing readers to the current publication schedule proposed by NIST [2]. Agencies planning for adoption had to plan for uncertainty about timing as well as content.

The comment period closed on September 12, 2017, for a draft that reorganized the catalog and moved guidance out of it. The draft’s notes said NIST continued to work with the privacy community and was particularly interested in how best to integrate privacy and security controls [1]. That signaled the privacy structure itself was open.

Finally, the draft did not say how its privacy designations would work outside the federal structure. Appendix F tied privacy authority to OMB Circular A-130. A non-federal organization has no such circular. It had to decide for itself which office owned which control.

Objections, in order

1. A draft changes nothing until it is final

Legally, that is right. Agencies remained on Revision 4. But a draft from NIST signals direction, and organizations with long system lives plan for it. Requirements written into a 2017 contract might govern a system for years. The draft gave program offices notice that controls would be phrased and organized differently.

2. Removing the actor makes accountability vaguer

The concern is reasonable. A control without a named actor could let everyone assume someone else does it. The draft answered by requiring organizations to designate responsibility. In our reading, the vagueness was already present when “the organization” stood for whoever happened to operate a system. The new phrasing made the assignment visible as a decision that must be recorded.

3. Non-federal organizations already had their own frameworks

They did, and the draft did not displace them. Its contribution for those organizations was a detailed control library that could sit behind a higher-level framework. The notes said the revision promoted integration with other approaches, naming the Cybersecurity Framework [1]. A wider audience did not require those organizations to switch.

4. Integrating privacy burdens security teams

Integration put privacy controls in the same document security teams read. The draft did not make security teams own them. Appendix F gave privacy programs sole authority over privacy controls and shared authority over joint ones. The burden fell on coordination, not on transferring work. That coordination cost is real, but it existed before integration too.

5. A catalog without baselines is unusable

The 2017 draft kept its baselines. The objection applies more to the direction of travel than to the draft itself. Even there, separating baselines from controls does not remove baselines. It puts them in a document that can change on its own schedule. Organizations that want a starting point still have one.

6. The changes were editorial

One could read the outcome rewrite as wording only. The capabilities stayed the same; the sentences lost their subject. But wording decides what an assessor tests and what a contract requires. A control that says the system must do something cannot be met by a manual procedure without an argument. A control with no actor can. That difference has consequences for design, staffing and evidence.

What an agency CISO or system owner could check

The draft’s changes are now settled in the final catalog. The checks they suggest still apply.

  • For each system, list every selected control and confirm the plan names who implements it, including cloud and service providers.
  • Check that every inherited control points to the common control provider and to that provider’s current assessment evidence.
  • Review organization-defined parameters and confirm they match current policy and contract terms.
  • Confirm which controls the privacy program owns and which are shared, and that both programs agree.
  • Read the latest system security plan for tailoring decisions without recorded rationale.
  • Check that solicitations for new systems state control requirements, including developer-facing ones, before award.
  • For any mapping from controls to a legal obligation, confirm someone has checked the obligation itself, as the draft warned.

Each check tests the layer the 2017 draft moved the decisions into: assignment, inheritance, parameters, ownership and rationale. None of them can be answered by reading the catalog. They can only be answered from an organization’s own plans, contracts and assessment records. In our reading, that is where the draft relocated accountability, and where an assessor will look first.

For a non-federal organization using the catalog voluntarily, the same checks apply with one addition. It should write down why it chose its baseline or overlay, since no federal policy will supply that reason for it.

With hindsight

The schedule in the draft’s notes did not hold. NIST published a final public draft of Revision 5 in March 2020, and the comment period was extended to May 29, 2020 [4]. The final Revision 5 followed in September 2020. Its NIST page records that a version dated September 23, 2020 was superseded by an update that includes changes as of December 10, 2020 [6].

The final version completed the separation the draft had begun. Its list of significant changes included removing control baselines and tailoring guidance from the publication and transferring the content to SP 800-53B [5]. That companion, Control Baselines for Information Systems and Organizations, provides three security baselines, one for each impact level, and a privacy baseline applied irrespective of impact level [7].

The privacy baseline was a change from the draft. In 2017, privacy-related controls were not allocated to baselines and were selected through Appendix F’s tags. In the final arrangement, SP 800-53B gives privacy its own baseline. Our reading is that the final version kept the draft’s idea of a privacy selection logic separate from impact level, and expressed it as a baseline.

The families changed too. The final catalog lists twenty families that include PII Processing and Transparency and a new Supply Chain Risk Management family. The draft’s Individual Participation and Privacy Authorization families do not appear in that list [5]. The final’s list of changes also named establishing a new supply chain risk management control family [5].

The applicability language changed to match the title. The final version says it establishes controls for systems and organizations and that the controls can be implemented within any organization or system that handles information. It says their use is mandatory for federal information systems under OMB Circular A-130 and FISMA [5]. The tension the draft carried between its title and its applicability section was resolved.

The final version also described the outcome-based change in the terms this article uses. It said controls were made more outcome-based by removing the entity responsible for satisfying the control from the control statement [5]. In our reading, that confirms the 2017 draft’s structural change survived intact.

Revision 4 was withdrawn on September 23, 2021, one year after Revision 5 was published, according to its NIST page [3]. The catalog then kept changing in smaller steps. On August 27, 2025, NIST issued a minor release, Release 5.2.0. It added control SA-24 and enhancements SA-15(13) and SI-02(07), and revised SI-07(12) and the discussion of several acquisition and integrity controls [6].

Several of the 5.2.0 changes fell in the System and Services Acquisition family that the 2017 draft had singled out as directed at developers. NIST’s page also now offers mappings from Revision 5 to the Cybersecurity Framework, the Privacy Framework and ISO/IEC 27001:2022, and the controls in a machine-readable format [6]. SP 800-53B received a matching 5.2.0 release with no changes to the baselines [7].

What held up is the frame. The catalog speaks to any organization, phrases controls as outcomes, holds privacy alongside security and leaves selection to a separate document. What did not hold up is the timetable and some of the family structure. The question the draft raised, who selects and who implements each control, is now answered in plans and contracts rather than in the catalog itself.

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

  • NIST SP 800-53
  • security controls
  • privacy controls
  • FISMA
  • control baselines
  • system security plans