The primary responsibility of the derivative classifier is to determine whether the new product incorporates classified information and, if it does, to ensure that information is properly marked, controlled, and protected at the correct level. The derivative classifier owns the classification decision on the new item, not the people who requested it, printed it, or plan to share it.

That answer has a boundary worth stating up front. Derivative classification applies to information the classifier can trace back to an authorized source document, or to a properly guided compilation of unclassified items whose combined meaning crosses a classification threshold. If the new product pulls from a classified hard drive copy that has no classification guidance attached, the classifier cannot responsibly skip the derivation step just because the material looks familiar.

The wrinkle readers rarely hear about: derivative classification is a formal authority, not an informal favor. Under the U.S. executive branch system, a properly trained and authorized derivative classifier can be anyone the original classification authority has empowered through agency training programs, including contractors. The original classification authority (OCA) is the only person who can decide that information is classified in the first place. The derivative classifier can only repackage, restate, paraphrase, compile, or incorporate decisions someone else already made.

That distinction carries real consequences. When a derivative classifier gets it wrong in either direction, the failure mode differs. Over-classification buries material behind controls that cost money and slow work. Under-classification exposes sources, methods, or programs. Both are reportable events in most agency programs, and both trace back to the same missed question: did anyone actually check what the source said before the new product went out?

The specific figures that govern this work live in the national policy documents. Executive Order 13526, the current governing executive order for classified national security information, establishes that derivative classifiers must carry training and must mark products with the source of the classification. 32 CFR Part 2001, the Information Security Oversight Office's implementing directive, sets the minimum training requirements and the agency-level program structure. Those two documents anchor nearly every agency's internal derivative classification guide, and they define the limits of what a derivative classifier may and may not do.

A practical example makes the stakes concrete. Suppose a program analyst assembles a briefing slide from three source documents marked SECRET, plus a screenshot of an unclassified news article. The analyst is not the one who classified the three sources. The analyst is not the one who decides what the news article "really means." The analyst's job is to apply each source's classification markings to the portion of the slide where that material appears, apply the correct overall banner, and document the derivative source on the slide or in an accompanying record. If the analyst instead eyeballs the slide and stamps it SECRET because "it feels sensitive," the analyst has stepped outside the authority the program granted, and the resulting mark is unsupported.

Who actually performs that derivation varies widely across agencies. In the Department of Defense, derivative classifier training flows through the Defense Counterintelligence and Security Agency (DSCA) and each component's security office. In the intelligence community, individual agencies run their own programs, often with additional controls on compilation and foreign disclosure. Contractors get derivative classification authority through their facility security officer and the contracting activity's security office. The training requirement is the same across all of them: every derivative classifier must complete training before applying classification markings, and must be re-trained on a recurring schedule that each agency sets.

The timing of that retraining matters for the workflow described above. Most agencies require derivative classifier training annually or biennially, and a training lapse suspends derivative authority until it is restored. A classifier who lets the certification expire cannot legally stamp a new product, regardless of experience. That creates a specific operational failure: teams with a single trained derivative classifier suddenly cannot release any derivative product when that person is on leave, transferred, or lapsed. Good programs train multiple people per work center for exactly this reason.

Here is where the exercise gets less clean. Compilation is the trickiest derivative classification path because no single source document sits behind the finished product. A derivative classifier can compile unclassified items into a classified product only when the combination reveals something the individual items do not — for example, a list of port visits whose pattern reveals an operational schedule. That judgment call requires the classifier to understand what the compilation reveals, not just what each item says. Most agency guides require a second reviewer or a formal compilation review for this reason.

Foreign disclosure adds another layer. Even when a derivative classifier correctly marks a product, sharing it with a foreign partner may require separate authorization under the National Disclosure Policy. That is not the derivative classifier's decision to make in most cases; it belongs to the agency's disclosure office. The classifier's responsibility stops at ensuring the product is correctly marked for its current authorized audience and that any foreign disclosure caveats carried by the source are preserved.

Let me put the core responsibility into a one-line working rule. A derivative classifier's primary job is to reproduce the classification decisions of the original classification authorities faithfully, not to make new ones. Everything else in the work — the markings, the source records, the training, the reviews — serves that single reproduction duty.

Here is where the training records come in for the audit question. A derivative classifier's source documents, training completion records, and marked products are typically retained for inspection by the agency's security office and by the Information Security Oversight Office's oversight program. In DoD, that inspection authority runs through DSCA. A missing source record on a classified product is a finding, even if the classification level happens to be correct. The record is what separates a documented derivative decision from an unsupported impression.

Time to compute the cost of getting this wrong. Under 32 CFR 2001, each improper classification is a reportable event in many agency programs, and corrections typically require a declassification review rather than a simple sticker change. A product marked at the wrong level may have to be recalled, re-reviewed, and re-released, and the labor hours involved run from a few to dozens per item depending on how many recipients already hold copies. The less obvious cost: over-classified products go through mandatory declassification review at the 25-year mark under Executive Order 13526, and each improperly marked item adds work to that backlog. Over-classification is not free protection; it accrues future cost.

That reframes the whole practice. The point is not that a derivative classifier should err on the safe side by classifying more. The point is that the classifier should classify exactly what the sources say, no more and no less. "When in doubt, classify higher" is a common folk rule, and it conflicts with the policy preference for the least restrictive classification that protects the information. The government's regulatory framework treats over-classification as a real failure, not a harmless instinct, because over-classification impedes information sharing, slows coalition operations, and costs public money in declassification review.

A classifier who follows that rule faithfully will often face pressure to go the other way. Program managers sometimes push to declassify a product to share it with a partner, a vendor, or a media team. The classifier's defense against that pressure is a specific document trail: here is the source, here is the marking, here is the training record, here is the review. If someone wants to release the material at a lower level, the request goes to the original classification authority or its designated declassification authority. The derivative classifier does not lower the level unilaterally, because that would mean making a new classification decision — which is not the derivative classifier's authority.

Now to distinguish this from the terminology readers often confuse it with. The term "original classifier" refers to the OCA who made the initial decision. The term "authorized holder" refers to anyone with lawful access to the information; that person can read it, but cannot classify anything. The term "derivative classifier" is the role with the delegated marking authority. The most common misuse is calling any security officer a "classifier," which flattens three distinct roles into one. A security officer who has not completed derivative classifier training cannot lawfully mark a new product, even if that officer handles classified material daily.

Twenty-five years is the trigger for automatic declassification review under the executive order, with specific exemptions for material that agencies have formally moved to a later date. That timeline is the long shadow of derivative classification decisions: a slide marked SECRET this year will be reviewed under that 25-year clock, and the source record attached at creation is what makes the review tractable. Derivative classifiers who skip the source citation work are, in effect, writing a puzzle for a reviewer decades from now.

To illustrate how the whole chain works, consider a training office that prepares a classified course module under the Modernized Integrated Database (MIDB) framework used across the Department of Defense. The course author pulls operating doctrine from a SECRET field manual, a map from an unclassified atlas, and a schedule pulled from a classified database. The derivative classifier certifies that the field manual material carries the source marking, that the schedule material is properly attributed to its classified source, and that the atlas content stays unclassified. The classifier then sets the overall module marking at SECRET with a "derived from" line naming the sources. The author cannot skip the classifier step; the classifier cannot skip the source check.

The specific documents a derivative classifier relies on are usually three: the source document with its original classification markings, the agency or program classification guide that covers the topic, and the training record that proves the classifier is current. If any of those three is missing, the correct move is to pause and route the question to the agency security office, not to guess. That backup step is the single most consequential procedural habit for a derivative classifier, because it converts an uncertain judgment into a routed question with a paper trail.

Where the training policy differs by agency is instructive. Some agencies require a formal derivative classifier course plus an annual refresher; others roll the training into general security awareness training and require a separate certification before actual marking authority. The common thread across all of them: nobody exercises derivative authority without documented training, and the certification is verifiable on request. A classifier who cannot produce that certification on request cannot defend the mark.

One failure scenario worth spelling out: an authorized holder copies a classified briefing into a personal shared drive so the team can work on it. The team edits the copy and releases the new version without a derivative classifier review. The release is now a derivative classification event that occurred without a derivative classifier. That is a reportable security incident at most agencies, even though no one had harmful intent. The primary responsibility exists precisely to prevent that unmarked drift from becoming an uncontrolled release.

The interviews with security officers across DoD work centers tend to surface the same practical tool: a derivative classification worksheet that records the source, the marking applied, and the classifier's name and certification date alongside the product. That worksheet is not a legal mandate under the executive order, but most component security offices require some equivalent. The worksheet makes the classifier's reasoning auditable and makes the source lookups fast for any later reviewer.

Tying the responsibility back to the executive order, the specific language in Executive Order 13526, Section 1.6, addresses derivative classification directly. It states that persons who reproduce, extract, or summarize classified information, or who apply classification markings derived from source material or a classification guide, must possess the authority to do so and must mark the product consistently with the source. That section is the textual anchor for everything above. A derivative classifier who can quote it — even roughly — wins nearly every internal dispute about who was supposed to check what.

The last practical point is the simplest. A derivative classifier's primary responsibility is not to classify the product at a level that feels right. It is to classify the product at the level the sources show, with the source record attached, and to route anything uncertain to the authority that can actually decide. That standard is achievable on any given Tuesday and auditable a quarter-century later.

How Derivative Classification Differs From Original Classification

The most important line to hold in this work is the boundary between the person who decides that information is classified and the person who applies that decision. The first role belongs to the original classification authority, an official formally designated in writing by an agency head or senior official under Executive Order 13526. The OCA can classify information in the first place and can authorize a classification guide that defines categories and levels. The second role, the derivative classifier, can only apply decisions the OCA already made.

That difference shapes every daily task. An OCA deciding whether a new program budget line is classified is exercising original authority, and the decision is subject to OCA review and reporting. A derivative classifier marking a briefing slide that restates the budget line is exercising delegated authority, and the correct behavior is fidelity to the OCA's decision, including its level and any dissemination controls. The slide inherits the marking; it does not improve on it.

The confusion shows up most often when a program creates a new product that synthesizes material across sources. An analyst summarizing a classified memo and an unclassified news article into a single brief may be tempted to "just mark it all SECRET." That move improperly escalates the unclassified content to a controlled marking, which creates unnecessary handling restrictions and can trip declassification review years later. The correct move is to mark each portion according to its source and to set the overall marking per the guidance for the highest source present.

The practical distinction pays off during reviews. Original classification decisions require a specific rationale, a classification duration, and an OCA identity. Derivative decisions require a source, a marking consistent with the source, and a trained classifier's name. Reviewers can check a derivative product in minutes against its source, which is the point of the structure. Original decisions require a different review because they involve a new judgment; derivative decisions require a fidelity check because the judgment already exists.

What the Source Documents Require Before Any Marking

A derivative classifier's work begins before any mark is placed, and the first step is confirming what the source actually says. The classifier needs the source in its original or properly authenticated form, with the classification markings and any portion markings intact. A screenshot, a recollection, or a verbal summary from a colleague does not satisfy that requirement in most agency programs. The source record is the classifier's working material.

Completing derivative classifier training is a hard precondition, and most agencies refuse to accept a marked product from a classifier whose certification has lapsed. The required interval varies: many components require annual refresher training, and the Defense Counterintelligence and Security Agency (DSCA) maintains the training schema for DoD components. Contractors get their training through their facility security officer and the contracting activity's security office. The certification date is the gate, not the years of experience.

A classification guide covering the subject matter is the third piece. Guides are issued by the original classification authority and define what categories of information are classified at which levels and for how long. When a classifier's product falls squarely under a guide provision, the guide supplies the level and duration. When the product does not fit any guide provision, the correct move is to route the question to the agency security office rather than to improvise.

If any of those three elements is missing, the classifier's correct response is to pause. A paused product with a documented routing question is a controlled situation. A product released without the three elements is an uncontrolled one, and that is exactly the situation the derivative classifier is supposed to prevent. The pause is the professional move, not a sign of ineptitude.

Marking the New Product: Portion Marks, Banner, and Source Line

The mechanics of marking a new product follow a predictable sequence under 32 CFR 2001 and the executive order. Portion markings come first: each paragraph, slide, or image carrying classified content gets its level marked at the location. The overall banner marking follows, set at the highest level present in the product. A "derived from" line naming the source or classification guide follows the banner. Declassification instructions or a review date complete the marking set when the source carries them.

The order matters because it forces the classifier to think in portions rather than totals. A product with three paragraphs, two of which restate SECRET portions and one of which restates an UNCLASS one, should carry SECRET portion marks on two paragraphs and an UNCLASS mark on the third. The overall banner is SECRET. A reader can then see which content drove the level, and a reviewer can check the marks against the sources quickly. Products marked with only a banner and no portion marks are harder to evaluate and more likely to over-classify.

The "derived from" line is where many production workflows break down. On a physical product, the line sits below the banner. On an electronic product or a slide deck, the line sits on each slide or in a document properties field. The line names the source document, the classification guide, or both. It is the audit trail that proves the mark was carried forward from a source and not invented at the release desk.

An important nuance in the regulatory text: the mark's level must be consistent with the source, but the mark cannot disclose the source beyond what the source allows. That means a product may need a marking that does not reveal which specific source document drove the level. In those cases, the classifier uses the guide's provision or a generic source citation approved by the security office. The classifier's job is to satisfy both the marking requirement and the source-protection requirement at once, and when those two conflict, the security office resolves the conflict.

Compilation, Paraphrase, and the Judgment Calls That Trip People Up

Compilation is the derivative classification path with the least obvious source record and the most potential for error. The rule under 32 CFR 2001 is that a compilation of unclassified items can be classified when the compilation reveals something the individual items do not. A list of location visits, a pattern of procurement dates, or an assembled organizational chart can fall into that category. The classifier has to evaluate what the compilation reveals, not merely sum what each item says.

That judgment requires the classifier to understand the operational context. A derivative classifier who can see the compilation's pattern but cannot articulate why it is sensitive should route the question rather than guess. Many agency programs require a documented compilation review by a second reviewer for exactly this reason. The review is the safeguard against both under-classification and the reflexive over-classification that "sounds safe."

Paraphrase is a subtler tripwire. A paraphrase that restates the same information is a derivative product, not a new invention, and the marking follows the source. A paraphrase that strips context and changes meaning is a new problem: the classifier cannot mark a distorted restatement as if it faithfully carried the source's content. The rule is simple to state and hard to apply: the paraphrase must convey the same information at the same level as the source.

The failed paraphrase is the classic incident. A writer edits a classified memo into a public-affairs style summary, drops the qualifying language that narrowed its meaning, and a reviewer marks the summary as a faithful restatement. The result can release more than the original did even when the letter of the marking is preserved. Good derivative classifiers spot that drift and send the product back for correction, which is why the job calls for reading skill as much as policy knowledge.

Training Currency and What a Lapse Actually Stops

Derivative classifier authority is time-bound. Training lapses suspend the authority, and the suspension is functional, not symbolic. A classifier whose card expired last quarter cannot lawfully mark a new product this quarter, and any product marked in that window lacks a valid source-of-authority record. Agency security offices routinely audit for this, and a found lapse can trigger a review of every product the classifier handled in the gap.

The lapse risk argues for training more than one person per work center. Teams that rely on a single trained derivative classifier see their release schedule stall whenever that person takes leave or changes assignments. The fix is straightforward: cross-train at least one backup per shop. Programs that skip that step discover the lapse when a product is due to a senior official and no one with authority is available to sign it.

Contractor handling differs in the source of the certification, not in the substance of the requirement. Contractor personnel get derivative classifier training through their facility security officer, and their authority flows from the contracting activity's security office rather than directly from the agency. When a contractor's certification lapses, the contract's security requirements typically pause classified deliverable marking until the training is restored. The administrative steps go through the FSO, not the employee alone.

There is also a public-facing pressure to consider. A lapse is discoverable through records and can become part of an office's inspection findings, along with the products affected. The reputational cost of that finding is real, and it is why most security offices send renewal reminders months before expiry rather than at the deadline. The deadline is unforgiving in a way that the calendar reminder is not.

Over-Classification vs. Under-Classification: Real Costs, Real Failure Modes

Both error directions are failures under the executive order's framework, and the framework does not treat them symmetrically. Under-classification exposes sources, methods, or programs to unauthorized disclosure, which is a direct harm and sometimes a permanent one. Over-classification imposes real but recoverable costs: restricted access, an expanding declassification backlog, and slowed information sharing with partners and the public. Neither is acceptable, and neither is safe as a habit.

The executive order's stated preference is for the least restrictive classification that protects the information, which is why "classify up when in doubt" is folk medicine rather than policy. Over-classification is tracked as a compliance problem in many agency programs and shows up in inspection findings. The 25-year automatic review under Executive Order 13526 makes the long-term cost visible: every over-marked item enters the declassification workload, and the aggregate cost of that backlog falls on the same agencies that created it.

Under-classification, by contrast, can trigger incident reporting, notification to affected parties, and corrective action requirements. The breach's scale depends on the product's distribution, which the derivative classifier often cannot know after the fact. That unknown is one reason agencies demand the source record attached at creation: it speeds the response when a marking problem is discovered later.

The working rule that reconciles those two costs is fidelity to the source. A classifier who marks exactly what the sources show, with the source record attached, is defensible in either direction of error. A classifier who substitutes instinct for the source is not, because there is no trail to check and no basis to correct the mark. The record is the defense.

Foreign Disclosure and Dissemination Controls Ride Along With the Mark

Derivative classification and foreign disclosure are adjacent but separate processes. A product marked correctly under derivative rules may still carry dissemination controls (such as NOFORN, REL TO, or controlled unclassified information designations) that limit who can receive it. Those controls come from the source and must be preserved by the derivative classifier. Releasing the product to a foreign partner without the disclosure authority's approval is a separate failure from misclassification, and it can be the more serious one.

The National Disclosure Policy governs that release process, and the disclosure office, not the derivative classifier, makes the call on foreign release. A derivative classifier's job is to make sure the markings and dissemination controls from the source are carried forward faithfully. That includes the source's own foreign-relations caveats, which sometimes appear as portion markers and sometimes as overall controls. Losing them in a derivative product is a common and reportable error.

There is a practical overlap worth knowing. A product that mixes classified portions with unclassified ones can sometimes be released at the lower level after a proper declassification review of the classified portions. That review belongs to the declassification authority, and the derivative classifier's role is to prepare the product for that review with clear portion marks. A product with clean portion marks makes the declassification review fast; a product with only a banner and no portion marks makes it slow or impossible.

The disclosure office also interprets partner-by-partner rules, which vary by country and by agreement. Derivative classifiers should not attempt to substitute their own judgment about partner eligibility. The safe path is to preserve controls, prepare clean portion marks, and route the release question to the disclosure office. That routing is not bureaucratic delay; it is what prevents an avoidable incident.

Common Misconceptions and Where the Practical Guidance Goes Wrong

One persistent misconception is that any security-cleared person can mark a classified product if the sources are in hand. The training requirement and the documented authority say otherwise. A cleared analyst with years of classified exposure but no current derivative classifier certification cannot lawfully place the marks, and products marked that way lack the authority record a review will demand. Cleared access and derivative authority are different statuses.

Another misconception is that derivative classification is only about labels. In practice, the classifier's most consequential work is the judgment about what the source says and how the new product aligns with it. The mark is the visible output, but the source analysis and the route to the security office when a question arises are the substance. A product with an impeccable banner and no source check can be wrong in ways the mark will never reveal.

A third misconception holds that all derivative decisions should "default to higher." That instinct fails under the executive order's preference for the least restrictive level and shows up as a compliance problem in its own right. Over-classification affects sharing, slows foreign disclosure, and adds declassification workload that lands on the agency's books for decades. The correct default is the source's level, not a personal safety margin.

The last misconception is that the derivative classifier's decisions cannot be reviewed. They can be, and they typically are: by the security office, by the inspection program, and by the declassification reviewers decades later. The record of source and training is what makes each of those reviews possible. A derivative classifier who cannot produce that record on request has done the marking without the accountability that the program requires.

FAQ

What single responsibility sums up the derivative classifier's job?

Reproducing the original classification authority's decisions faithfully, with the source record attached, sums it up. The derivative classifier does not make new classification decisions; the person applies the level and duration the source or classification guide already set. That limit is the practical definition and the one that holds up in a review.

Can a contractor serve as a derivative classifier?

Yes, provided the contractor completes the required training and receives the delegated authority through the contracting activity's security office. Contractor certification flows through the facility security officer, and the certification date is what the security office checks. The substance of the duty is the same as for a government employee: apply the source's decision, not a new one.

What happens if a product is released without derivative classifier review?

It is typically treated as a reportable security incident, even without harmful intent. The incident response usually involves recalling the product, reviewing the markings against the sources, and correcting the release. The person responsible for the office's security program is the one who records and routes the incident, so early reporting matters more than internal debate.

Does the 25-year review apply to every derivative product?

Only to material the executive order covers, and only to portions not covered by a later, properly approved declassification exemption. Executive Order 13526 sets automatic review at 25 years for classified records meeting the order's scope, with agency-specific exemptions for certain categories. The source record attached at creation is what makes the later review tractable.

Is over-classification really a problem if the material is sensitive?

Yes, because the executive order's framework prefers the least restrictive level that protects the information, and over-classification produces its own compliance findings. It also adds declassification workload, slows foreign disclosure, and restricts sharing with cleared partners who could otherwise use the material. The correct level is what the source shows, not a margin above it.

What is the difference between a derivative classifier and an original classification authority?

The OCA decides whether new information is classified; the derivative classifier applies the OCA's decision to new products. The OCA's decision comes with a rationale, a duration, and a named official. The derivative classifier's output carries the source, the marking, and the classifier's own certification date, and does not require a new rationale.

Can an analyst mark a product if a derivative classifier is unavailable?

No, unless the analyst personally holds current derivative classifier certification. Cleared access is not the same as delegated marking authority, and the training record is what the security office checks when a marking is questioned. The safe behavior when the appointed classifier is unavailable is to route the product to a backup or pause release until the authority is present.

How is a compilation handled if no single source document exists?

The classifier evaluates whether the combination reveals something the individual items do not, and many programs require a second reviewer for that judgment. If the compilation crosses a classification threshold, the product is marked with the appropriate source or guide citation and the review record is kept. If the classifier cannot articulate the revealed information clearly, the question routes to the agency security office before any mark is placed.

The material above is general educational information about classification program practice and is not a legal opinion. Specific classification decisions remain with the original classification authority, declassification authority, or disclosure office designated by the relevant agency. For a live classification question, consult your agency's security office or the contracting activity's security representative before marking or releasing any product.