You've probably seen the stamp. REDACTED. CLASSIFIED. TOP SECRET. Now, it shows up in movies, news leaks, and FOIA requests with thick black bars across entire paragraphs. But here's the thing most people never stop to ask: who actually decides what gets that stamp? And who decides which stamp it gets?
It's not some algorithm. On the flip side, it's not a committee that meets in a basement. It's a specific person — or a specific role — with specific authority granted by the President. And the rules around it are stricter than most people realize.
What Is Classification Authority
Classification authority is the legal power to look at a piece of information and say: this stays secret. Not just "keep it quiet" — but legally secret, backed by criminal penalties for unauthorized disclosure. That authority doesn't float around loosely. It flows from the top down It's one of those things that adds up..
Here's the thing about the President holds original classification authority by virtue of the office. That's it. On top of that, the Constitution makes the President the head of the executive branch, and classified information lives almost entirely inside the executive branch. So the President is the source Simple as that..
But the President doesn't sit around marking documents. They're doing derivative classification. Everyone else? Through Executive Order 13526 (signed by Obama in 2009, still in force), the President designates certain officials as Original Classification Authorities — OCAs for short. These are the only people who can originally classify information. Day to day, they delegate. Big difference Simple, but easy to overlook..
Original Classification Authority
An OCA is a senior official — think agency heads, deputy secretaries, certain military commanders — who has been formally designated in writing. That said, a CIA deputy director might have Top Secret authority for intelligence sources and methods. A Navy admiral might have Secret authority for submarine capabilities. Here's the thing — the designation has to specify exactly what topics they can classify and at what level. They can't just classify anything they feel like Worth knowing..
And they can't delegate their OCA status. They can authorize others to derivatively classify based on their guidance, but the power to make the original call? That stays with the named OCA.
Derivative Classification
It's where most classification actually happens. On the flip side, a derivative classifier isn't making a new judgment about whether something should be secret. They're applying existing classification guidance to new material. In practice, say an OCA writes a classification guide saying: "All technical specifications for the X-47B drone are SECRET. Plus, " An engineer writing a maintenance manual for the X-47B? Here's the thing — they're a derivative classifier. Day to day, they mark the manual SECRET because the guide says so. Even so, they don't decide if it's secret. They just follow the rules.
Derivative classifiers still need training. They still need authorization. But they're not OCAs. The distinction matters — legally and practically.
Why It Matters
Over-classification is a real problem. The government classifies millions of documents a year. Some estimates put it over 50 million classification decisions annually. A huge chunk of that is derivative — people marking things secret because the guide says so, or because "better safe than sorry," or because they genuinely don't know what's already public.
But under-classification is dangerous too. In real terms, if an OCA fails to classify something that should be protected — a covert officer's name, a technical vulnerability in a weapons system — people can die. Now, operations fail. Advantages evaporate.
The system tries to balance both risks. Executive Order 13526 requires that classification meet specific standards: the information must be owned by, produced by, or for the U.government; unauthorized disclosure must reasonably be expected to cause damage to national security; and the classifier must be able to identify the specific category of damage (military plans, intelligence activities, foreign relations, etc.S. ) Easy to understand, harder to ignore. Simple as that..
It also requires a classification reason — a short code like "1.Think about it: 4(e)" for scientific/technological matters. 4(c)" for intelligence activities or "1.And a declassification instruction — usually a date (25 years is standard) or an event ("declassify on death of source").
None of this is optional. If an OCA skips the reason or the declassification instruction, the classification is technically invalid. That doesn't mean you can ignore it — but it means it can be challenged It's one of those things that adds up..
How Classification Decisions Actually Work
Let's walk through what happens when a new piece of sensitive information lands on someone's desk.
The OCA Decision Process
An Original Classification Authority receives or generates information. Maybe it's a new intelligence report. Which means a weapons test result. A diplomatic cable.
- Is this official government information? Personal opinions, pure speculation, already-public facts — none of that qualifies.
- Does unauthorized disclosure reasonably threaten national security? Not "could embarrass us." Not "might be awkward." Reasonably expected to cause damage. That's a legal standard.
- Which category does it fall into? The executive order lists eight: military plans, WMDs, intelligence activities, foreign relations, scientific/technological, cryptology, vulnerabilities, and nuclear.
If the answer to all three is yes, the OCA assigns a level: Top Secret (exceptionally grave damage), Secret (serious damage), or Confidential (damage). Then they mark it, date it, sign it (or use a digital equivalent), and include the reason and declassification instruction And that's really what it comes down to. And it works..
That's the clean version. In practice? Consider this: they rely on classification guides — detailed documents they've already approved — to handle routine decisions. OCAs are busy. They only personally review the edge cases.
Classification Guides
A classification guide is the OCA's force multiplier. Day to day, it's a document that says: "For this program, these types of information are classified at this level, for this reason, with this declassification instruction. " It lets derivative classifiers work without calling the OCA every time Most people skip this — try not to..
Good guides are specific. Plus, bad guides say things like "all program information is SECRET. Here's the thing — " That's lazy — and technically non-compliant. The executive order requires guides to be precise enough that a trained derivative classifier can apply them without guessing.
Guides also have to be reviewed. At least every five years. In real terms, more often if the program changes. An outdated guide is worse than no guide — it produces systematically wrong classifications.
The Classification Levels in Practice
Top Secret — Reserved for the crown jewels. War plans. Nuclear codes. Identities of deep-cover agents. Penetrations of foreign governments. Disclosure = exceptionally grave damage. Access requires a Top Secret clearance and a need-to-know and often a special access program (SAP) or SCI (Sensitive Compartmented Information) indoctrination.
Secret — The workhorse level. Most military operations, intelligence reports, diplomatic negotiations, weapons capabilities. Disclosure = serious damage. Requires Secret clearance and need-to-know Worth knowing..
Confidential — The floor. Still legally protected, but disclosure = damage (not serious, not grave). Administrative data, some logistics, certain training materials. Requ
ired clearance and need-to-know.
The Derivative Classifier
Most classifications happen not in the OCA's office, but in the hands of derivative classifiers—engineers, analysts, contractors, even junior analysts—who apply established guidelines to specific documents. They're the backbone of the system, but they operate under strict constraints.
A derivative classifier can only classify information that's derivative—meaning it's based on or derived from already-classified sources. They cannot classify raw intelligence intercepts or original source materials—that remains the OCA's domain. More importantly, they must have the proper training and authorization to work within specific classification guides.
This creates a chain of responsibility. If a derivative classifier misapplies a guide, they're accountable. Even so, if an OCA's guide is flawed, that's a systemic failure. The system assumes competence at every level, but it's built on the premise that each link can be audited and corrected That alone is useful..
Time Sensitivity and Automatic Declassification
Classification isn't eternal. Think about it: every classified document has a built-in expiration date, though it's rarely "never. " The executive order mandates automatic declassification unless the OCA actively extends it. This creates a crucial feedback loop: information that remains sensitive must justify its continued secrecy, while information that loses its protective value sees it automatically returned to the public domain That's the whole idea..
This changes depending on context. Keep that in mind.
This process isn't smooth in practice. Here's the thing — agencies develop their own rhythms for review cycles, and some documents slip through the cracks—either declassified too early or held far beyond their justified lifespan. The system relies on both automated triggers and human judgment, which means inconsistency is built into the architecture.
The Human Element
Behind every classification decision sits a human judgment call, even when automated systems flag potential violations. OCAs don't just apply rules—they weigh consequences, consider context, and make calls about what the nation can afford to have public. This is where institutional memory matters, where understanding of programs and their vulnerabilities informs decisions about what truly constitutes a threat.
The system works because it balances legal precision with practical flexibility. It acknowledges that perfect consistency across thousands of classification decisions is impossible, but it creates enough structure and accountability that the inevitable errors can be identified and corrected Less friction, more output..
Conclusion
The classification system isn't airtight—it's functional. It trades absolute security for manageable risk, relying on trained professionals working within defined frameworks to make thousands of micro-decisions about what stays secret and what emerges. Think about it: the key insight is that classification is fundamentally about harm prevention, not information control for its own sake. Every document carries a justification for its secrecy, and that justification must survive scrutiny from multiple angles.
In the end, the system succeeds not because it's perfect, but because it's designed to evolve. Guidelines get updated, reviewers get refreshed, and the automatic declassification mechanisms make sure yesterday's secrets don't become tomorrow's unnecessary burdens. The real test isn't whether every classification decision is flawless, but whether the framework consistently produces more accurate judgments than it leaves uncorrected.
