What Is a Primary Responsibility of the Derivative Classifier When Creating a New Product From a Classified Source?

When creating a new product from a classified source, a primary responsibility of the derivative classifier is to determine whether the new material incorporates or reveals classified information and, if it does, to mark and control it at the appropriate level. That determination is an affirmative, document-by-document line-by-line review, not a clerical stamp inherited from the source. The derivative classifier owns the classification decision for the derivative, and that ownership cannot be delegated to the original author, a template, or a colleague's earlier markings.

Derivative classification sits at the junction of two separate authorities. The original classification authority (OCA) decides what is classified and why. The derivative classifier decides what in a new product carries that classification forward. Confusing the two roles produces both over-marking and under-marking, and both are security failures with different costs. This explainer covers the mechanism, the quantities involved, the edge cases practitioners actually hit, the misconceptions that cause most errors, the practical steps that hold up under review, and the limits of what any classifier can promise.

Mechanism: How Derivative Classification Actually Works

Derivative classification is the act of incorporating, paraphrasing, restating, or generating in new form information that is already classified, then marking the new material consistently with the source. The governing rule is simple in statement and demanding in execution: the derivative classifier must determine that the information is classified, that the source marking applies, and that the derivative's own markings are accurate and complete. Each of those three determinations is a separate failure point.

The mechanism depends on what the classifier can see and cite. A proper derivative decision traces back to a specific classified source, ideally one that already carries classification guidance such as a marking, a classification guide entry, or a security classification guide (SCG) provision. The classifier reads the source, identifies which portions are classified, decides which of those portions appear in the new product, and carries the level and category forward. The new product then needs its own overall marking, portion markings where required, and any dissemination controls the source imposed.

Three properties make this harder than it looks. First, aggregation: individually unclassified facts can combine into a classified picture, so a new product can become classified even when no single sentence matches a source. Second, reformulation: paraphrasing, summarizing, and translating are all covered, so a classifier cannot escape the obligation by rewording. Third, category transfer: special handling controls such as dissemination restrictions or controlled unclassified designations travel with the information, and dropping them is itself a violation even when the level is correct.

A useful mental model is that the classifier is rebuilding the classification from evidence. The source is not a rubber stamp; it is a citation. The output is a new classification decision that must stand on its own if reviewed years later by someone who never saw the drafting process.

Quantified Context: Scale and Review Burdens

The derivative problem is a volume problem before it is a judgment problem. Large agencies routinely process derivative products in the thousands per year, and each one carries a review obligation proportional to its length. A 40-page report with paragraph-level portion marking means dozens to hundreds of marking decisions, and a single missed portion can drive a re-review of the entire document.

Review budgets are asymmetric. The classifier's time per document is bounded by staffing, while the cost of an error is not. A mis-marked derivative can require recall, re-review, and notification, and the effort scales with distribution rather than with the drafting time. This is why experienced programs push classification decisions as early as possible, ideally before drafting, when the source set is small and the new product does not yet exist.

Comparative markings carry their own arithmetic. When a derivative draws on two sources at different levels, the derivative takes the higher level for the portions that carry both, and portion markings must reflect the actual combination rather than the dominant source. A common quantitative error is assuming one source's overall marking applies uniformly to every paragraph when the source itself carries portion-level exceptions.

Special handling adds a second count. Each dissemination control is a separate condition that must be evaluated per portion, so a document with two classification levels and three handling controls can require up to six distinct decisions per paragraph. Programs that track these as a single "classification" checkbox are the ones most likely to lose a control during reformatting or republication.

Edge Cases and the Counterintuitive Detail

The counterintuitive detail is that removing classified content does not automatically make a derivative unclassified. Because aggregation works in both directions, a new product that omits a classified detail but preserves the surrounding analytical structure can still reveal classified conclusions, force dispositions, or source vulnerabilities. A document can be classified because of what it implies, not only because of what it states, so a "clean" rewrite built from a classified source still needs the same review as a verbatim extract.

Several edge cases recur:

Paraphrase laundering: rewording a classified sentence and treating the result as new, unclassified analysis, when the underlying statement is still classified. Aggregation across an unclassified boundary: combining open-source material with classified fragments to produce a product that is classified as a whole. Foreign disclosure: producing a derivative for release to a foreign partner without evaluating the source's releasability controls. Republication drift: copying markings from an earlier derivative rather than the authoritative source, which propagates any error in the intermediate product. Template inheritance: using a previously approved template whose markings predate a change in the source's classification guidance. Automated generation: routing classified source text through a tool that summarizes it, then treating the output as unclassified because no human typed the classified words.

Each of these fails the same test. The derivative classifier must be able to point to a classified source and explain why the specific content in the new product is or is not classified. If that explanation is unavailable, the decision is not done.

Misconceptions That Cause Most Errors

The most widespread misconception is that classification is a property of the sentence rather than of the information. Sentences change; information does not. A classifier who marks only the sentences that look classified will miss paraphrases, summaries, and structured data that carry the same information in different form.

The second misconception is that the original classifier's markings are automatically correct for the derivative. Markings are context-dependent. A portion marked at one level in an operational cable may not need the same level in a strategic assessment, and a portion that was unclassified in one product may be classified when combined with new material. The derivative decision must be made in the derivative's own context, with the source as evidence rather than as instruction.

A third misconception is that the derivative classifier can approve new classification. Derivative classifiers cannot create classification where none exists; they can only carry forward what an OCA or a classification guide already established. If a new product contains genuinely new information that appears to warrant classification, the correct action is referral to an OCA, not an independent decision.

A fourth is that the classifier's job ends at the overall marking. Portion markings, dissemination controls, declassification instructions, and the source citation itself are all part of the derivative decision, and incomplete markings are treated as classification errors even when the level is right.

Comparison of Derivative-Classifiers and Related Roles

The distinctions matter in practice. A derivative classifier who behaves like an OCA creates unauthorized classification, which is its own compliance problem. A derivative classifier who behaves like a reviewing official skips the review, which shifts risk to whoever signs the product. The correct posture is narrow: apply existing classification accurately, cite the source, and refer anything outside that scope.

Timeline Table: Derivative Classification in the Product Lifecycle

The timeline shows why the earliest stages carry the most leverage. A classification decision made during drafting costs minutes; the same decision made after distribution costs recall and re-review. Programs that front-load the review consistently report fewer corrections, though the numbers vary too widely by agency to generalize.

Practical Actions: What a Derivative Classifier Should Do

The practical steps below are ordered by leverage, not by administrative convenience.

Build the source set before drafting. Identify every classified source the product may draw on and obtain the authoritative marking or classification guide entry for each. Record the source, the date, and the specific provision relied on. Decide classification before writing. Draft a short classification plan that lists the source content expected to appear and the level and controls each portion will carry. Revise the plan when the scope changes, not after the draft is complete. Review line by line against the sources, not against memory. For each portion, ask whether the content incorporates, paraphrases, restates, or reveals classified information, and mark accordingly. Treat aggregation explicitly: ask whether the combination of unclassified portions creates a classified picture. Transfer every control, not just the level. Dissemination controls, handling caveats, and declassification instructions apply to the derivative to the same extent they applied to the source. A correct level with a dropped control is still a derivative classification error. Cite the source on the derivative. The citation is the audit trail that lets a later reviewer reconstruct the decision without the drafting context. Derivatives without a source trail are the hardest to defend and the most likely to be returned for correction. Refer out what you cannot decide. New information that appears to warrant classification goes to an OCA. Foreign release questions go to the disclosure reviewer. A classifier who resolves these independently is acting outside the role.

A red-flag list is useful during review. Treat any of the following as a stop-and-recheck condition:

A derivative with no identifiable classified source citation. Portion markings that are all identical despite a source with portion-level variation. A product that draws on two sources at different levels but carries a single overall marking with no portion-level breakdown. Any reformatted, summarized, or machine-generated content whose markings were inherited rather than re-derived. Dissemination controls present on the source but absent on the derivative. A template or earlier derivative used as the marking reference instead of the authoritative source.

Limitations and What the Role Cannot Guarantee

Derivative classification reduces risk; it does not eliminate it. The decision depends on the quality of the source markings, and mis-marked sources propagate errors that no derivative review can detect without independent access to the original classification authority's records. That dependency is structural, not a training gap.

The role also cannot keep pace with aggregation at scale. A classifier can reason about the classified picture created by a handful of portions, but a product that combines hundreds of unclassified fragments from classified sources can produce a classified conclusion that no line-by-line review surfaces. Programs that routinely generate large analytical products need aggregation review as a separate step, not as an extension of routine derivative marking.

The final limit is temporal. Classification guidance changes, and a derivative that was correctly marked when produced can become incorrectly marked when the source is updated. Without a review cycle tied to source changes, derivative accuracy decays. No derivative classifier can promise that a product will remain correctly marked indefinitely; the realistic promise is that the marking was correct against the sources available at the time of review, and that the citation trail makes later correction possible.

This article is general educational information about classification practice and compliance roles, not legal advice, a security determination, or an official interpretation of any agency's classification guidance. Classification rules vary by jurisdiction, agency, and program, and readers should consult their organization's security office or a qualified professional for decisions about specific documents.