History and culture December 28, 2021 Updated: September 13, 2026 9 min read Urh Zagradišnik

Secrecy, Classification and Declassification: How the State Produces Secrets

A state secret is not the natural condition of information but an administrative decision with real consequences for public oversight. When does secrecy legitimately protect people and operations — and when does it begin protecting an institution from accountability?

A state secret is not the natural condition of information. It is produced by a decision: an official with authority judges that disclosure could cause sufficient harm and therefore restricts access. That decision can be reasonable. The identity of a living clandestine source, an active military plan, a cryptographic capability or a precise vulnerability in critical infrastructure are not things a responsible government should publish without thought.

But precisely because classification creates a severe information asymmetry, it is dangerous in its own right. The state can tell the public that it may not see something, while the public often lacks the information needed to test whether the restriction was justified. Secrecy is therefore both a protective measure and power over the evidentiary field. Whoever decides what remains hidden can influence what citizens, journalists, historians and even other parts of government are able to know.

U.S. Executive Order 13526 acknowledges this tension with unusual clarity: information may not be classified to conceal violations of law, inefficiency or administrative error, prevent embarrassment, restrain competition, or delay release of information that does not require national-security protection. If a system must expressly prohibit these abuses, secrecy cannot honestly be treated as a morally neutral label.

this article therefore does not begin with the naïve demand that everything must be public. It begins with a stricter question: when does secrecy genuinely protect people and legitimate defensive capacity — and when does it begin to protect an institution from accountability? Under the Natural Law lens used by THY-REALITY, a lawful stamp does not by itself create a moral right to hide the truth from those in whose name power is exercised.

A secret is not a property of a document but a decision of power

A classified document is not necessarily true, accurate or important. The marking primarily describes how information must be handled and what kind of harm its unauthorized disclosure is expected to cause. A poor analysis can be Top Secret; a public report can be rigorously evidenced. Classification status is therefore not a measure of truth and should never substitute for evaluating evidence.

Under U.S. Executive Order 13526, information may be originally classified only by an authorized official, when it is under government control, falls within defined security categories and disclosure can reasonably be expected to cause describable damage to national security. The system uses Confidential, Secret and Top Secret according to the expected severity of harm. If there is significant doubt about whether classification is needed, the order says the information shall not be classified; if there is doubt about the level, the lower level is to be used.

The United Kingdom uses a different administrative model built around OFFICIAL, SECRET and TOP SECRET. The comparison matters because it shows that secrecy is not a metaphysical category. States construct rules, thresholds and procedures. These differ, so a marking cannot be understood apart from the legal and institutional system that produced it.

What is legitimately worth protecting?

There are forms of information for which absolute openness would be irresponsible. Revealing the identity of an active human source can cost a life. Publishing a cryptographic vulnerability can enable intrusion. Details of a still-valid military plan, protective procedure or critical-infrastructure weakness can increase another actor's capacity to cause harm. EO 13526 therefore permits classification categories covering military plans, intelligence sources and methods, cryptology, certain foreign-relations information, system vulnerabilities and information concerning weapons of mass destruction.

This boundary is essential even for a system-critical publication. If every secret is declared an abuse, we lose the ability to distinguish protection of a person from protection of an institution. Natural Law does not require suicidal transparency. It does require proportionality: secrecy must protect a concrete legitimate good, not merely the convenience of power.

The real question is therefore not 'may a state have secrets?' but 'what harm is concrete enough to justify temporarily excluding the public, who makes that judgment, and who can review it?'. Without those three questions, national security easily becomes a phrase that grants itself permission.

Who decides — and how secrecy reproduces itself

Original classification is only the beginning. Much of the system operates through derivative classification: an official creates a new document from an already classified source or classification guide and carries forward the relevant markings and declassification instructions. The derivative classifier does not need the same original authority as the official who created the secret in the first place.

This creates a bureaucratic copying effect. One original decision can appear in a memorandum, assessment, summary, slide deck, email and later report. Each copy is not a new grand secret, but the system produces an expanding body of records that inherit the restrictions of an earlier decision. EO 13526 therefore requires training for derivative classifiers and explicitly emphasizes avoiding overclassification.

This matters because secrecy is not only created in dramatic decisions by agency chiefs. It is often more mundane: a marking is carried forward because preserving a restriction is safer for the official than assuming the risk of removing it. Institutional incentives can be asymmetric — an improper disclosure produces visible punishment, while unnecessary secrecy much less often produces a personal consequence.

When secrecy begins protecting the institution

Some of the strongest criticism of overclassification has come not only from outsiders but from within the U.S. political system itself. The Commission on Protecting and Reducing Government Secrecy associated with Senator Daniel Patrick Moynihan described secrecy as a form of regulation and warned that excessive secrecy can weaken protection of the secrets that truly matter. Later congressional hearings treated overclassification as a problem of cost, accountability and effectiveness, not merely an abstract preference for openness.

EO 13526 therefore contains a striking prohibition: classification may not be used to conceal violations of law, inefficiency or administrative error, prevent embarrassment, restrain competition, or prevent or delay release of information that does not require national-security protection. The rule reads almost like a map of the system's predictable temptations.

But a prohibition on paper is not a guarantee in practice. Institutions have a natural tendency to protect reputation, internal process and earlier decisions. The legitimacy of secrecy therefore cannot rest on the goodwill of the same organization that benefits from withholding. It requires deadlines, review, appeals, archives and a real possibility of outside scrutiny.

FOIA: a right to records is not a right to every line

The U.S. Freedom of Information Act gives the public an important tool for accessing federal agency records, but it contains exemptions. Exemption 1 protects information properly classified for national defense or foreign policy. Other exemptions can protect information withheld by another statute, personal privacy, certain law-enforcement techniques or confidential commercial information.

A response is therefore not always a simple yes or no. A document can be released in part: some paragraphs remain visible while others are redacted. FOIA processing includes a requirement to consider reasonably segregable non-exempt information. The familiar black bars are the visible product of this conflict between access and withholding.

Researchers must be disciplined here. A redaction proves that content was withheld; it does not prove what the content says. Five blacked-out lines are not evidence of an assassination, surveillance program or political order unless another source supports that conclusion. Criticism of secrecy becomes weaker, not stronger, when a blank space is filled with the story we wanted to find.

Glomar: when even the existence of a record can be secret

In some cases, merely confirming that a record exists would reveal protected information. U.S. practice therefore permits an agency to neither confirm nor deny the existence of requested records — the so-called Glomar response. EO 13526 expressly allows such a response where the very existence or nonexistence of records is itself classified, and the Department of Justice continues to address Glomar responses in modern FOIA oversight.

The mechanism displays the extreme form of information asymmetry. A citizen may not even receive an answer to whether the subject of a request is documented. In a specific case that can make sense: if the government effectively revealed its clandestine sources merely by confirming that a particular kind of file exists on a named person, the acknowledgement itself could disclose sensitive information.

But that is exactly why Glomar cannot become mystical proof of whatever hidden claim we prefer. 'We can neither confirm nor deny' does not mean 'yes, it exists'. For a researcher it means looking for independent records, court filings, congressional findings, archival traces or later declassification.

Publicly released CIA document with extensive black redactions. — Redaction makes the this article evidence rule visible: a black bar proves that content was withheld, not what the hidden content says. A released or declassified document can still be only a partial view of the archive. Image: Central Intelligence Agency (CIA) / Wikimedia Commons Public domain — U.S. federal government

The mosaic effect: harmless pieces can form a sensitive whole

The security problem does not always sit in one explosive sentence. Multiple apparently harmless data points can combine into a pattern: the identity of a source, the location of a device, the scope of a collection capability, the timing of an operation or a technological limitation. EO 13526 therefore allows a compilation of individually unclassified items to be classified when the combination reveals an additional association or relationship that meets classification standards.

This is the logic commonly described as the mosaic problem. In CIA v. Sims, the U.S. Supreme Court emphasized broad protection for intelligence sources and methods and the possibility that seemingly insignificant details can help an adversary assemble a larger picture. The argument is not automatically false simply because the public has difficulty independently testing it.

Yet the mosaic argument is also convenient because it can expand almost without limit. If every fragment can hypothetically become part of a dangerous mosaic, nearly anything could be withheld. The requirement of describable harm, proportionality and independent review therefore remains essential. The possibility of combining information cannot become a universal license for darkness.

Declassification: a secret needs a life cycle

A secret that made sense in 1980 may not make sense in 2026. An operation ends, technology becomes obsolete, a source dies, alliances change and a military plan ceases to apply. A system that knows how to classify but not how to open records is not merely protecting information; it is manufacturing an accumulating darkness.

EO 13526 requires declassification when information no longer meets classification standards. At original classification, a date or event should be set; where that cannot be determined, the ordinary framework is 10 years, extendable to as much as 25 years when sensitivity requires it. Permanently valuable historical records more than 25 years old are generally subject to automatic declassification, with defined exceptions for matters such as human sources, cryptology, current vulnerabilities and certain military or diplomatic information. In special cases protection can extend to 50 or 75 years.

There is also mandatory declassification review for sufficiently specific requests and an appeal path to the Interagency Security Classification Appeals Panel. The National Declassification Center at the National Archives coordinates large interagency declassification work. This is part of legitimacy itself: a government that demands temporary secrecy must also maintain a mechanism by which secrecy ends.

An archive is not a complete image of the past

Declassified archives are extraordinarily valuable, but they are not a complete recording of the state. Some documents were never created. Others were destroyed under records schedules, lost, misfiled or remain protected. Some records contain equities from multiple agencies and require additional referral. Others are released only in redacted form.

The absence of a document is therefore not automatic proof that an event did not occur — but neither is it proof that the event did occur. This symmetry is essential. If we say 'there is no paper because they must have hidden it', we have created a claim that cannot be falsified. Every lack of evidence then becomes evidence of concealment.

A better method is to build levels of confidence. Do we have a direct document? Independent testimony? Multiple archives? A court record? A later official confirmation? Or only a gap? The temptation to fill gaps is greatest around secret institutions, which is precisely why evidentiary discipline must be stricter.

Natural Law: the state does not own the truth

Reasonable secrecy protects a concrete person, defensive capability or legitimate operation from concrete harm. Abused secrecy protects power from the consequences of its own actions. Both can carry the same TOP SECRET stamp. The moral difference lies not in the color of the marking but in purpose, proportionality, duration and reviewability.

Under the Natural Law lens used by THY-REALITY, government does not own the truth about its own conduct. Citizens may sometimes be temporarily excluded from operational details when disclosure would cause clear and disproportionate harm. They cannot be morally excluded from governmental accountability merely because the government itself declared its conduct secret.

The sharpest boundary is therefore simple: secrecy can be legitimate as a temporary shield against a real danger; it becomes illegitimate when it becomes a permanent shield against truth. A democracy that cannot look behind its own black bars does not control its secret system. The secret system controls the boundary of what the democracy is allowed to know.