The reform summary surfaced last week. The nearest option on it is a 90-day restart.

For two issues this brief has declined to write about the reform task force, because nothing had actually moved. Something has now moved. Federal News Network obtained the Department’s internal summary of the 997 responses to July’s reform request for information and published a two-part investigation on October 26 and 27. The summary sorts the available options by who can act and how long it takes. The nearest one is restarting Phase 2 within ninety days.

First, what this is and is not. It is not a decision, an announcement, a memorandum or a rule. It is the task force’s own written summary of what industry told it, together with options grouped by feasibility, produced in early September and obtained by a news organization. The Department declined to comment on fifteen questions put to it about the contents. Nothing on the list has been adopted, and the public recommendations still have no release date. Read it the way you would read a draft board paper that leaked: not as policy, but as the clearest available picture of which way policy is being pushed.

The structure is the useful part. The summary separates the options into three groups by authority. The first group is what the Chief Information Officer could do without any rulemaking at all: reduce the ambiguity created by assessors, prime contractors and program offices interpreting the same requirements differently; move the program to Revision 3 of NIST SP 800-171, a transition that has been on hold since the current Chief Information Officer took office in December; and improve the consistency of CUI marking. The second group is slower, costlier, or needs rulemaking: extend plan-of-action-and-milestones deadlines, publish templates and configuration settings so that testing is easier, and deal with the overlap between these requirements and other regimes such as ISO 27001. The third group is outside the Department’s gift entirely — grants, subsidies and tax credits, which it could advocate for but cannot create.

  • Within 90 days
    Restart Phase 2 with a narrower population while longer-term work goes through rulemaking. Work with the accreditation body and the training arm on more consistent guidance for assessors and consultants. Publish assessor availability statistics.
  • 6 to 12 months
    Issue explicit guidance on crediting comparable ISO 27001 requirements. Reduce the volume of CUI the government sends to primes and subcontractors in the first place.
  • A year or more
    Publish approved configurations and templates on the model of the Department’s hardening guides. Develop machine-readable evidence criteria using OSCAL. Offer free tools, subject to funding.
  • Through rulemaking
    Expand plan-of-action-and-milestones eligibility to 270 days, described in the summary as part of an amendment already in process.

Now the phrase that will be quoted out of context for the next six months. On the forthcoming FAR rule for controlled unclassified information, the summary says the existing draft “has the potential to drastically reduce the burden on industry” by “narrowing the population that must comply with NIST 800-171,” and that improving the clarity of CUI marking has “the greatest potential to narrow the DIB population subject to NIST 800-171 implementation.”

Read the mechanism rather than the adjective. The lever being described is marking — the government being more precise about what it hands you and what it calls controlled. A company leaves that population when the government stops designating its data as CUI, which is the fix industry has asked for since the program began and which this brief has covered as the single most-cited friction in the response record. It is not a lever that reduces what you owe if you do hold controlled information. And two items on the list move in the other direction: taking the Revision 3 transition off hold raises the baseline, and machine-readable evidence criteria raise what counts as having demonstrated it.

Ninety days from when, and why the range matters: the ninety days run from a decision nobody has announced. Katie Arrington, who performed the duties of the Department’s chief information officer and championed the program, expects the pause to run roughly two more years. So the honest range is three months to two years — and only one end of it carries a cost if you plan wrong, because a program that restarts on ninety days’ notice finds you exactly where you are.
So what: strip out the items that depend on a decision and look at what is left standing on every tier of that list. Consistent assessor interpretation, templates, ISO crediting, machine-readable evidence, a longer plan-of-action window — every single one of them is a change to how you demonstrate compliance, and each presumes you have a current system security plan, named evidence behind each requirement, and a clear view of which of your data is actually controlled. None of it is a discount. A longer plan-of-action window is more time to close a gap you have already written down and disclosed, not permission to have one. Crediting an ISO 27001 control is a mapping-and-evidence exercise, not a waiver. And if the Revision 3 transition really does come off hold without rulemaking, the version of the standard you are measured against changes before any assessment schedule does — so read it now rather than when it arrives.

The response data does not say what the pause said it said. Asked what costs the most, industry named the system security plan.

Part one of the same investigation, published October 26, works through the response data itself. The Department reviewed 997 responses. Just over half — 509 — came from defense industrial base companies; the rest came from vendors, consultants and others. The suspension in July was announced on two grounds: a shortage of third-party assessors, and compliance costs that were structurally incompatible with small business. The responses speak to both grounds, and not in the direction the announcement implied.

The stated reasonWhat the response data showsWhat it means for you
Assessor capacity was the binding constraintAround 47% of respondents cited assessor cost — fifth among cost drivers, not first. Capacity was reportedly never more than half used during Phase 1, and 2,433 companies held Level 2 certificates by September, roughly five times the Department’s own first-year projection. Average assessment: about $47,000.Treat the queue as available, not full. Capacity is not what will stop you, and the much larger cost figures in circulation do not come from this data set.
Certification cost is what is crushing small businessThe top cost driver named was implementing and maintaining the NIST SP 800-171 controls themselves — not the certification wrapped around them. Licensing, enclaves and inconsistent CUI marking followed. On overhead, respondents ranked system security plans highest and FIPS-validated encryption second.Your cost problem is the controls and the documentation. Neither a pause on verification nor a restart touches either one.
A 75-day review would settle itSixty days for feedback plus fifteen to write the report, now past a hundred days. Roughly twenty people worked through more than 1,100 comments in three days, largely by hand. The 2024 rulemaking took about eight months for roughly 1,400 comments.The delay is a throughput problem, not a signal that the obligation is being withdrawn. Do not read slowness as softening.

On whether the self-assessment regime does any good, the data is genuinely mixed and should be quoted carefully: 58% of those who answered said self-assessments improved their cybersecurity, and 43% called them a check-the-box exercise — but fewer than a fifth of respondents answered that question at all, so both figures are a minority view rather than a referendum. What respondents did not say, anywhere in the summary, is that the program’s burden was too high.

Eric Crusius of Hunton made the point that organizes all of this: the cost complaints are aimed at the certification program when what they actually describe is the underlying NIST requirements, which the certification program does not change. The reporting also notes that the current Chief Information Officer had met the program office for roughly twenty minutes since her confirmation in December, and that the program office was not consulted before the suspension — which is relevant here only for what it says about durability. An arrangement built without the people who run the thing is not the arrangement most likely to last.

So what: the request for information asked industry what costs the most and what carries the heaviest overhead, and on the second question the answer was the system security plan. That document is not an assessment artifact. It is a requirement in its own right under NIST SP 800-171, it is the basis on which your SPRS score is calculated, and it exists whether or not anybody ever comes to look at it. So price it honestly and then finish it — and make sure it describes the system you actually run, not the system you designed. If yours was written to support a score and has not been revised since, it is simultaneously the most expensive document you own and the least defensible. Second item on the same list: go and confirm that your cryptographic modules are FIPS-validated, with a certificate number, rather than merely described as AES-256 somewhere in a vendor datasheet.

NIST spent an hour last Thursday on how a requirement is scored. One rule in the slides explains how a self-scored 110 becomes a negative number.

On October 29, Victoria Pillitteri, a co-author of NIST SP 800-171, and Daniel Eliot, who leads NIST’s small business engagement, spent an hour walking through the assessment side of the standard. This brief covered the publication of their small-business primer in September; the webinar is where the mechanics got explained, and the slides are now public. They are worth an afternoon, because they describe the arithmetic that decides whether you are compliant, and it is not the arithmetic most self-assessors use.

A requirement is not scored as a requirement. It decomposes into assessment objectives — evaluation statements that break the requirement into separately testable parts. Each objective is examined, asked about, or tested, and not every method applies to every objective. Each one comes back either satisfied or other than satisfied. Then comes the rule that does all the work, stated plainly in the deck: every objective is binary, and if any single objective is not met, the entire requirement is Not Met. There is no partial credit at the requirement level.

That is why a requirement you are confident you meet in substance can fail on one line — the log retention period nobody set, the shared account nobody disabled, the procedure that exists but was never approved. And Revision 3 adds a second trap. It introduced organization-defined parameters, and the slides show, using device lock as the worked example, that the parameters are themselves assessment objectives. The organization has to have selected its own values — how long a device sits idle before it locks — before the remaining objectives can be assessed at all. A company that never wrote the number down has not implemented the requirement loosely. It has failed an objective, and the whole requirement with it.

Where the 280-point gaps come from: the public enforcement record includes a contractor that scored itself 110 and was later assessed at −170. Read against this deck, that stops being mysterious. A self-scorer asks “do we do multifactor authentication?” and answers yes, honestly. An assessor asks every objective underneath that requirement, finds one that is other than satisfied, and marks the whole requirement Not Met — zero credit, and then the point deduction. Repeat that across a hundred and ten requirements weighted at one, three and five points each, and a very large gap arrives without anyone having set out to misrepresent anything. The gap is usually not a lie. It is the difference between a question asked once and the same question asked every way it can be asked.

The rest of the deck is process, and one line in it is aimed at owners rather than engineers. The four phases are prepare — scope the assessment and gather the artifacts: plan of action, system security plan, network diagrams, policies — then develop the assessment plan, conduct it, and document and report. An external assessor normally writes that plan, and the slides tell business leaders to read it before it runs, for scope and for business disruption. Afterwards: one plan-of-action entry per Not Met requirement with corrective actions, resources and target dates; update the system security plan to the system as implemented; report the score; monitor continuously.

So what: do this with the five requirements that protect the most valuable thing you hold. Open SP 800-171A Revision 3, write out every assessment objective underneath each one, and mark it satisfied or other than satisfied with the evidence named beside it. Then confirm that every organization-defined parameter in your environment has a value written down where an assessor could read it. Note too where this meets the top story: machine-readable evidence criteria sit on the reform list as a one-year-plus item, and that is this same structure in a form a machine can check. Companies that can already answer at the objective level will find the transition administrative; companies that cannot will find it an audit.

What to do this week

Every option on the reform list assumes you already hold these.

  • Read your system security plan against the system you actually run. It has to describe the environment as implemented.
  • Decompose your five highest-value requirements into assessment objectives, and give every organization-defined parameter a written value. One objective other than satisfied fails the whole requirement.
  • Read Revision 3 now. Taking the transition off hold needs no rulemaking.

Quick hits

  • Late Oct 2026
    The assessment queue turned back up while the program requiring it is still suspended.
    A practitioner recap of the October accreditation-body town hall reports final Level 2 certificates up 8% since September and 184 Level 2 assessments in progress; the body has not published its own October figures, so treat these as indicative. The number worth noticing is 184. The in-progress queue stood at 151 at the end of September and had been falling for months, which read as a backlog draining. It is now rising — companies are booking third-party assessments they are not currently required to hold, during a suspension, with no restart date announced.
  • Ongoing
    There is a no-cost federal platform that builds the two documents this issue is about.
    Project Spectrum has been run by the Department’s Office of Small Business Programs since December 2019, and under Section 1807 of the FY2026 defense authorization act maintaining it is now a statutory obligation rather than a discretionary program. Registered small and medium businesses get cyber readiness checks, risk and vulnerability assessment, continuous threat monitoring, system security plan development and plan-of-action tooling. A current Department contract is not required — firms pursuing contracts, subcontracts or grants qualify, and 21,535 companies were participating as of the most recent published count. It lowers the cost of producing the artifacts; it does not sign your affirmation.

Not sure where your organization stands?

CUI Vault gives you a live, always-audit-ready picture of your CUI environment and NIST 800-171 posture — so headlines like today’s are a footnote, not a fire drill. Reach out to your NtelSec team to talk through your current gaps.

This briefing is a general-information summary of publicly reported CMMC and CUI developments. It is not legal advice; consult your compliance counsel before making contractual or certification decisions.
Talk to us about CUI Vault All issues