If you put a derivative classification on a document, the source material you're deriving from has to be analyzed just as carefully as the document you're marking. That means confirming the source is authorized, current, correctly marked, and actually contains the information your new document carries. Skip that review and you inherit every error the source holds β€” and the responsibility lands on you, under Executive Order 13526 and the implementing directive 32 CFR Part 2001.

Derivative classification is the everyday engine of the U.S. classification system. Most classified documents produced each year are not originally classified by an original classification authority; they carry markings copied or paraphrased from existing classified sources. Executive Order 13526, Section 1.1(a)(3), defines derivative classification as incorporating, paraphrasing, restating, or generating in new form information already classified, and applying the same level of classification to the new document. The Information Security Oversight Office (ISOO), which oversees the program through 32 CFR Part 2001, frames the derivative classifier's job as carrying forward decisions made by an original classification authority β€” not making new ones. The whole system rests on a simple premise: if the source is right, the copy is right. That premise fails the moment the source review is sloppy.

What Derivative Classification Actually Requires of You

The person applying derivative markings is not a passive stamper. Under the standard training regime established by ISOO, a derivative classifier must demonstrate knowledge of classification levels, the marking process, the classification authority block, the source document, the reasons behind the source's classification, and the prohibition on downgrading or declassifying inherited information. That's a longer list than most new classifiers expect. The source analysis step sits at the center of it. You need to read the portion markings on the source, check its authority block, and confirm the document you are producing is actually covered by that source. If you can't point to the specific paragraph or attachment that carries the classification, you don't have a source β€” you have a guess.

Time pressure is where this breaks down. A six-page derivative product built from a 200-page source package can look routine until you trace a paragraph marking back and find it was added by a secondary document that itself cites a source you can't access. The safe move at that point is to stop and ask. Ask the originator, ask the security officer, ask the original classification authority named in the block. Nobody penalizes a classifier for requesting confirmation. People get penalized for guessing.

Reading a Classification Authority Block Before You Copy Anything

The classification authority block is the paper trail, and it should tell you three things: who originally classified the information, what category of information it falls under, and when it can be downgraded or declassified. On a derivative document, the block normally names the source document rather than an individual official β€” the "derived from" line. Under 32 CFR 2001.22, that block also carries a declassification date or event. If either line is missing, unreadable, or internally contradictory (a date before the source document's own date, for example), the source has a defect and you should treat it that way.

Watch for composite documents. A briefing deck assembled from three different source products can carry three different authority lines, and a single "derived from" entry won't capture all of them. In that case, list each source or attach a classification guide reference. The 32 CFR 2001.22 structure allows either source documents or a security classification guide as the basis for derivation, and a mixed package usually needs both. If the deck conflicts with itself β€” one slide marked Confidential from one source, another marked Secret from another β€” the correct derived level is the highest one present, but your source listing still has to identify each contributing document.

Version drift is a common trap. Source documents get revised, and a classification that applied to revision A may not apply to revision C. If your source copy is more than a few years old and the program it belongs to has changed, confirm with the originating office that the marking still reflects the current decision. The classification authority, not the date on your file copy, controls.

The Difference Between Deriving From a Source and Relying on a Guide

Classifiers have two legitimate bases for derivative marking: a classified source document, or a security classification guide issued by an original classification authority. They are not interchangeable, and the distinction shapes how carefully you read. A source document tells you what was classified in that specific product. A guide tells you which categories of information are classified at what level and for how long, and it's designed to let you classify new material that didn't exist when the guide was written. If you're working from a guide, the analysis step is checking whether your specific content falls within a described category β€” not whether it resembles a paragraph you saw somewhere.

Basis for derivation What it contains What you must verify Typical failure
Classified source document The classified information as it appears in that product Portion markings, authority block, document currency Citing a superseded revision
Security classification guide Category descriptions, levels, declassification rules Whether your content matches a described category Stretching an unrelated category to cover new material
Composite/mixed package Multiple sources or guide plus document Every contributing source listed separately One "derived from" line masking three sources

Guides carry a trap of their own. Because they describe categories rather than specific sentences, a classifier can talk themselves into covering material the guide never contemplated. That's the failure mode to watch for. Apply the guide narrowly. If your new document asserts something the guide's category description doesn't reach, the content isn't derivatively classifiable from that guide, and it may need an original classification decision instead.

Portion Marking Analysis: Where Copy Errors Become Over-Classification

Portion markings are where derivative errors show up most often and where they cost the most. Under 32 CFR 2001.24, each portion of a classified document β€” each paragraph, title, subject line, and graphic β€” carries its own level marking, and the overall document marking is the highest portion level. A derivative classifier who marks the whole document Secret because one paragraph is Secret, without marking the unclassified paragraphs as unclassified, has produced a document that looks fully Secret. Over the following years, that document gets referenced, copied, and stored as Secret. The error compounds.

ISOO has repeatedly identified over-classification as one of the program's persistent problems, and the Public Interest Declassification Board has argued for years that the volume of over-marked material strains declassification capacity. The arithmetic is straightforward. If a 40-page derivative product has three classified paragraphs and 37 unclassified ones, and only the classified paragraphs are marked, every page reads as classified at storage and review time. Multiply that by every derivative product in a program and the declassification backlog grows for no reason.

The practical test is mechanical. Read the source portion by portion, and mark your output portion by portion at the same level. Where the source portion is unclassified, mark yours unclassified β€” "U" on every unmarked paragraph is acceptable, though many agencies use the convention of only marking classified portions plus a document-level note. Never combine portions at different levels into a single paragraph. If you're paraphrasing two paragraphs of different levels into one sentence, the sentence takes the higher level, and the source listing should reflect both.

Why the "Derived From" Line Is the Chain of Custody for Classification

A derivative document's "derived from" line works like a chain of custody. It lets anyone downstream β€” a reviewer, a declassification analyst, an inspector general β€” trace the marking back to a specific decision. When that line is vague ("multiple sources," "various briefings," or blank), the marking becomes unfalsifiable, and unfalsifiable markings tend to be wrong. The federal standard in 32 CFR 2001.22 asks for identification of the source document or classification guide, and ISOO's self-inspection guidance treats missing or generic source citations as a finding.

Trace one case through. An analyst writes a two-page summary of a program status. The summary marks two sentences Secret and cites "program office briefing" as the source. Ten years later, during a mandatory declassification review, nobody can locate the briefing, confirm its authority, or verify the level. The two sentences either get released because the chain is broken, or they get held because nobody can justify release. Both outcomes are bad. Both trace back to a one-line shortcut taken at drafting time.

The remedy is boring and effective. Name the specific document and date. Keep a digital copy or a retrievable reference number. When a guide is the basis, cite the guide's title and date. When there are multiple sources, list each. Ten extra seconds at drafting time saves hours of downstream review, and reviewers can act on a well-cited marking without calling anyone.

Common Failures: Stale Sources, Composite Documents, and Inherited Guesses

Three failure patterns account for most derivative classification errors. The first is the stale source: a document marked under a classification guide that has since been revised, or a source whose declassification date has already passed. A derivative product carrying expired restrictions is misclassified by definition, and the classifier who copied it forward owns the mistake unless the source marking itself was obviously contradictory.

The second is the composite document problem, described above. The third is the hardest to catch: the inherited guess. Somewhere in the source package, an earlier classifier marked something classified without a clear basis, and that marking has been carried forward through several derivative generations. By the time it reaches you, the original basis is three documents back. The authority block may cite a document that cites a document that cites a guess. This is the scenario where asking the original classification authority is not optional.

Consider a mid-size agency processing a contract deliverable that cites a prior deliverable, which cites a program office memo, which cites an "OGC review." The OGC review turns out to have addressed a different question entirely. The classification was applied by analogy, not by authority. A derivative classifier who reads the chain recognizes the gap; one who stops at the immediate source does not. Reading one link back is easy. Reading two or three links back is what separates careful analysis from copying.

The Personal Accountability That Follows a Marking You Apply

Sanctions for misclassification apply to the person who applied the marking, not only to the original classifier. Executive Order 13526 and its implementing directive allow for sanctions against employees who knowingly or negligently classify improperly, and agency self-inspection programs sample derivative products specifically to find markings that can't be traced. The practical exposure has two sides. Over-classification is the more common finding; under-classification, when someone copies a lower level by mistake or misses a portion marking, is the more serious one.

Negligence is the standard most often cited in training, and it's the reason the analysis step matters. A classifier who demonstrably checked the source, documented it, and asked when uncertain has a defensible record. A classifier who stamped a document from a source they hadn't opened does not. The record is the source citation and the paragraph-level trace, which is exactly why the authority block and portion markings deserve the attention they get.

Derivative classifiers also can't downgrade or declassify what they derive from. If a source says Secret, your product says Secret unless an authorized authority has changed the decision. Classifiers sometimes think a short excerpt or a paraphrase is "less sensitive" and drop a level. That's not a judgment call available to them. If you believe the source marking is wrong, the path is a challenge to the original classification authority, documented and processed β€” not a lower marking applied quietly.

What a Document Review Looks Like When the Source Is a Guide Only

Working from a guide shifts the analysis from copying to matching, and the sequence is different. A guide-based review has to establish that the guide is current, that your content falls inside a described category, and that the level and duration you apply match the guide exactly. That's three checks, and each one fails in its own way. An expired guide is common after program reorganizations. Category mismatches show up when analysts stretch a definition. Wrong durations appear when someone applies a level correctly but a declassification rule from memory.

A practical sequence for a guide-only document: pull the guide and confirm its effective date and issuing authority; identify the category descriptions that plausibly cover your content; map each portion of your draft to exactly one category or to none; apply the level and duration from the guide without adjustment; record the guide's title, date, and the category numbers applied. If a portion maps to no category, leave it unclassified and note why. If it maps to two categories at different levels, apply the higher level and list both categories.

Guides also change the declassification line. A source document carries a date tied to that document. A guide carries a rule β€” often an event, like "25 years from the date of the source document" or a specific event tied to a program milestone. Your derivative product's declassification line follows the guide's rule, not the date you happened to draft. Getting this wrong delays release by years and creates the kind of backlog the Public Interest Declassification Board has flagged repeatedly.

Balancing Under-Classification Risk Against Over-Classification Cost

The analysis step is often described as a hunt for over-classification, but the greater risk runs the other way. Under-classification can expose sources and methods, and the sanctions regime treats it as the more serious error. A classifier worried about speed should worry more about missing a portion marking than about marking one level too high. The asymmetry is real and agencies say so in training. One is a compliance problem. The other is a compromise.

Error type Typical cause Detection path Practical consequence
Over-classification Copying the highest level to the whole document Self-inspection sampling; declassification review Information withheld past need; review backlog grows
Under-classification Missed portion marking or lower-level paraphrase Source comparison during review Potential unauthorized disclosure; sanctions exposure
Untraceable marking Vague "derived from" line Review of authority block Marking cannot be defended or released cleanly

Over-classification still carries costs, and they're easier to measure than they used to be. Every unnecessarily marked paragraph has to be reviewed later. Multiply 37 unclassified pages wrongly marked Secret across a program's output and the review hours add up fast β€” hours that could have been spent on the paragraphs that were actually sensitive. The careful review step cuts both directions. It catches over-marking as reliably as it catches under-marking, provided the reviewer actually reads the source.

The middle path is documentation. Mark at the level the source supports, cite the source precisely, note any judgment you made and why, and escalate ambiguity. A defensible marking is one a reviewer can follow from output back to authority in three steps or fewer.

Building a Repeatable Source-Review Habit for Derivative Work

A repeating habit beats a heroic one-off review. The habit has four moves and takes minutes per document once it's routine. First, open the source. Not the citation β€” the actual source document or guide. Second, read the authority block and the declassification line before reading the body. Third, map each portion of your draft to a portion of the source, marking at the source's level. Fourth, record the source by name, date, and category, and note any portion you couldn't map.

1: Open the cited source and confirm it is the current version. 2: Read the classification authority block and declassification line. 3: Match each draft portion to a source portion or guide category. 4: Apply the source's level to each matched portion; mark unmatched portions unclassified. 5: Record source name, date, and category numbers in the authority block. 6: Escalate unmatched or contradictory content to the security officer or original classification authority.

When step 3 fails β€” a paragraph in your draft has no counterpart in the source β€” the failure-handling matters more than the marking. Don't invent a level. Pull the paragraph out, mark it pending review, and route the question to the security officer with the specific language at issue. The same applies when the source's own markings conflict. A document that says Confidential in the header and Secret in the body needs review before it becomes a source, not after.

Teams that run this sequence consistently report fewer corrections at release time, because the errors surface while the drafter still has context. The habit also builds a searchable trail. Six months later, when someone asks why a paragraph is marked Secret, the answer is in the block, not in a memory.

FAQ

Is derivative classification different from original classification?

Yes, and the difference is who makes the decision. An original classification authority decides that specific information requires protection; a derivative classifier only applies that existing decision to a new document. ISOO training materials put the distinction plainly: derivative classifiers carry forward decisions, they don't create them, and they have no authority to change a level, add a category, or shorten a declassification date.

Can I downgrade or declassify a document I'm deriving?

No. The implementing directive bars derivative classifiers from downgrading or declassifying inherited information. If the source is Secret, your product is Secret until an authorized official changes the source decision. A challenge to the original classification authority is the only route, and it should be documented in writing rather than handled through a lower marking.

What happens if I can't find the source document cited in the block?

Treat the marking as untraceable and escalate before applying it. An untraceable source is the classic inherited-guess scenario, and the safe action is asking the security officer or the named original classification authority for confirmation. Applying a level you can't substantiate creates a record you can't defend during review, inspection, or litigation.

Does a declassification date in the past mean the document is automatically public?

Not automatically, but it does mean the marking is out of date and the document should have entered review. A derivative product built from a source past its declassification date carries an expired restriction, and the correct action is referral to the agency's declassification authority rather than copying the expired line. Some categories are exempt from automatic declassification under 32 CFR 2001.26, so the past date requires confirmation, not assumption.

How long does a careful source review take per document?

For a short derivative product citing one source, the review is typically a matter of minutes: open the source, check the block, match portions, record the citation. Composite products citing three or more sources take longer because each contributing document needs its own check. The time cost sits well below the cost of a correction cycle after release.

Is a security classification guide enough by itself as a source?

Only if your content actually matches a category the guide describes. Guides are authorized bases for derivative marking under 32 CFR 2001, but they're written at the category level, and classifiers sometimes stretch a category to cover material it never contemplated. Match narrowly, and if the content falls outside every category, it isn't derivatively classifiable from that guide.

What should the "derived from" line actually say?

Name the specific source document or guide, with a date and, where applicable, a document number or reference. A generic entry like "multiple sources" or "program office briefing" fails the traceability standard in 32 CFR 2001.22 and is a typical self-inspection finding. When several sources contribute, list each separately or attach a source annex to the classification authority block.

Does careful analysis only matter for paper documents?

No. The rule applies identically to briefing decks, emails marked classified, database records, and any other format where derivative markings are applied. Digital products are often harder to review because portion markings get stripped during format conversion, and a slide deck that loses its paragraph-level markings in export becomes uniformly classified by appearance. Check the output format before release.

This article is general educational information about classification practice, not legal advice or an agency-specific determination. Classification rules, guide contents, and declassification exemptions vary by agency and by the governing directive in force at the time. Consult your agency's security officer or original classification authority for decisions about specific documents, and follow your local reporting channels for any suspected misclassification.