Criminal Justice & the Rule of Law

Can Trump Just Declare Nuclear Secrets Unclassified?

Alex Wellerstein
Thursday, August 18, 2022, 8:01 AM

It’s not entirely clear whether the former president can use an appeal to presidential authority to declassify any secrets, but in the case of “Restricted Data,” the classification category for nuclear secrets, it seems unlikely.

Former President Donald Trump speaking at CPAC (Gage Skidmore,; CC BY-SA 2.0,

Published by The Lawfare Institute
in Cooperation With

One of the apparent defenses offered up by former President Trump’s lawyers to the discovery of highly classified materials at Mar-a-Lago is that Trump had a “standing order” that any classified materials he took home with him were, by fiat, declassified. Never mind for a moment about whether there is any record of this order from when Trump was president or whether he had the putative power to issue it. Never mind, either, whether this means that these documents are now unclassified and should be available to anyone willing to file a Freedom of Information Act request for them. Let us instead look at whether any president actually has the power to do this, especially given that these documents allegedly contain papers related to nuclear weapons secrets.

National Defense Information

Most official secrets in the United States are what are known as National Defense Information, which includes anything classified as “Confidential,” or “Secret,” or “Top Secret,” and so on. The origins of this system date back to World War II, which precipitated the need to combine three pillars of governance into one secrecy regime. 

The first pillar comprises presidential executive orders that govern what kinds of classification categories there are, what they are supposed to be applied to, and who can apply them or declassify them. The first of these was issued by Franklin Roosevelt in 1940. Executive Order 8381, “Defining Certain Vital Military and Naval Installations and Equipment,” indicated that certain categories of information (for example, “All official military or naval books, pamphlets, documents, reports, drawings, photographs, contracts or specifications”) could be marked “secret,” “confidential,” and “restricted.” Over the years, additional executive orders have added categories (“Top Secret” was added in 1942) and defined what fell under them. For example, “Top Secret” was defined as “information the security aspect of which is paramount and whose unauthorized disclosure would cause exceptionally grave danger to the nation.” Then President Barack Obama issued the most recent of these executive orders in December 2009. Executive Order 13526 outlined who can classify documents, what kinds of classification categories can exist, and what procedures exist to declassify documents.

The second pillar of the American secrecy regime includes the regulations that federal agencies promulgate to enact the demands of these executive orders. For example, a Top Secret document will have certain handling requirements that a Secret document does not. In the past, one could send a Secret document by registered mail in some circumstances, whereas a Top Secret document would require an armed courier to transport it. Some kinds of documents require storage in a safe of a given rating. Others can be photocopied only on machines whose innards have been properly vetted to make sure they are not capable of surreptitiously storing copies. The details in some of these regulations are mind-bogglingly specific, because this is how you turn a broad goal like “keep this piece of paper secret” into a specific reality. 

The third pillar is the law: the statutes that tell you what happens to people who break the aforementioned regulations. Chief among these is the Espionage Act of 1917, which details possible punishments for people who disclose or mishandle national security information (as defined by the executive orders), depending on their intent and other considerations. The law, in this case, doesn’t tell you much about the content of the secrets: By itself, it is just the “teeth” that enforce the previous two pillars, which define the secrets and the implications of them.  

So, could Trump declassify National Defense Information at will? Essentially, the logic is that whoever has the authority to classify documents also has the authority to declassify documents. And, all classification authority under these executive orders ultimately stems from the president, so the assertion that Trump could have declassified those documents as he saw fit is probably plausible. However, valid declassification actions require procedures, records, and logging, and arguably ought to have been marked as declassified prior to entering into private hands. This defense has never been litigated, however, so the limits a president might have in this area remain uncertain. Typically, classification violations are handled administratively—revoked security clearance, terminated employment, etc.—and not as criminal cases.

Restricted Data

National Defense Information covers pretty much every secret you can imagine, except one special category: anything relating to the production or use of nuclear weapons. This type of information has its own special legal framework that is parallel to and independent of all other secrets, although any given nuclear secret might (and frequently does) exist in both regimes simultaneously (requiring two sets of clearances, for example, and two different declassification actions, one for each system). 

In the months after the end of World War II, members of Congress debated about what kind of legal framework should govern the newly created, nation-spanning nuclear weapons infrastructure that had been erected in secret by the Manhattan Project. The “battle of the atom,” as one congressman called it, was a question about what kind of atomic future the country, and perhaps the world, might have. 

One of the core debates was about secrecy. Scientists associated with the Manhattan Project had testified at length—and even hired the Ad Council to push the message—that there was actually “no secret” to making an atomic bomb (the full mantra was “no secret, no defense, international control,” the latter referring to plans for a treaty to ban nuclear weapons). They created their own newsletters and organizations to lobby for what they saw as a more sane law. Atomic bombs were products of scientific research that had been widely published before World War II. The discovery of nuclear fission and nuclear chain reactions was no secret. Unlike, say, the specific military plans for D-Day, atomic secrets were in a sense “public,” because anyone with a properly outfitted and funded laboratory could discover them. So, they argued, trying to control the spread of nuclear weapons through secrecy was going to be not only futile (because the Soviets, and others, also had good scientists) but also counterproductive, because excessive secrecy would hinder domestic activity in the field as well. Science, they argued, was an international activity that needed circulation of ideas and data to progress. Secrecy would, they predicted, strangle its advancement in any country that tried to implement it. 

The first draft of the Atomic Energy Act was sponsored by the War Department and would have given the newly created Atomic Energy Commission the power to issue regulations controlling sensitive nuclear information. Many American scientists, including many former Manhattan Project scientists, feared that this would lead to an ever-expanding sphere of secrecy, and lobbied surprisingly successfully to kill the bill (an event so rare that it was given a name: the Scientists’ Movement). The next offering, created by sympathizers to the scientists after long hearings of their grievances, was known as the McMahon Act, after Sen. Brien McMahon, D-Conn., a junior senator who was chairing the Special Committee on Atomic Energy. The McMahon Act’s initial draft had essentially no secrecy in it: Its main provision about information was about dissemination and indicated that nuclear weapons would be treated like any other military technology in that the underlying concepts and science would not be classified, but specific applications might be. The initial draft proposed that any classified information would just fall under the Espionage Act. In other words, the idea of a tank wouldn’t be classified, but the specifics of an M-1 Abrams might be. 

This pleased the scientists but made the military queasy. They felt that the Espionage Act might not be adequate for atomic secrets because it did not protect “information” per se, and they generally disagreed that the distinction between basic science and applied technology would work in practice. Gen. Leslie Groves, head of the Manhattan Project, leaked information to the press about a minor Soviet spy ring in Canada in early 1946, which set off the first round of atomic spy fears, and resulted in the McMahon committee revising their bill significantly. The final version of the McMahon Act had removed the section on “dissemination of information” and replaced it with a section dedicated to “control of information,” which included an expansive new secrecy category and huge penalties for violating it.

This new category, called “Restricted Data,” was defined as “all data concerning the manufacture or utilization of atomic weapons, the production of fissionable material, or the use of fissionable material in the production fo power.” Restricted Data was regulated by the Atomic Energy Commission (AEC). The AEC could remove information from the category if it determined the information could be published “without adversely affecting the common defense and security,” but it could not increase the scope of what fell under the definition. This is a very counterintuitive way to think about secrecy. The information is made secret based on its inherent semantic content, not by a classifier. The role of the AEC classification expert here is not declaring a document to be Restricted Data, but identifying Restricted Data within the document. 

However strange, this legal construction was deliberate. The idea was to simultaneously have wide-ranging secrecy in the area of nuclear weapons—and, initially, nuclear power—while avoiding the possibility that the AEC could run amok with its secrecy powers. The AEC “can only reduce—it can not enlarge—the scope of the crime,” as McMahon put it. So, Congress enlarged the extent of secrecy to its maximum by default and allowed the AEC only to reduce it.

Over the years, the “all data” part would be taken very literally: All data relating to atomic weapons, irrespective of who discovered or invented it, or whether they worked for the U.S. government or not, would be covered. For this reason, Restricted Data is frequently referred to as “born secret,” because the classification requirements technically apply the moment the information itself comes into existence, however it does. I am not a government scientist, but if I were to come up with a nuclear weapons design on my own, it would, according to the Atomic Energy Act, need to be regulated. If I didn’t regulate it appropriately, I could go to jail, pay a fine, or (until 1969) even be executed. The Atomic Energy Act was given a significant overhaul in 1954, and has had modifications since then, but this provision  remains on the books today.

But, is this really enforceable? Surely this “born secret” interpretation of the clause would quickly run into the First Amendment when applied to private citizens? The truth is, nobody really knows. The AEC and its successor organizations—most notably the Department of Energy—have generally avoided trying to find out if a court would actually back them up on this interpretation of the law, even though they have publicly and privately stated that this is also their interpretation of the law. In practice, the “born secret” approach has been used to encourage voluntary compliance—from journalists, academic scientists, and industry—but not as an actual punishment. For would-be atomic spies, the Espionage Act has appeared to be entirely adequate. For example, the Rosenbergs were convicted of conspiracy to commit espionage under it, rather than for the distribution of Restricted Data under the Atomic Energy Act.

There has been only one case in which the “born secret” provision of the Atomic Energy Act was actively tested in court, which I detail in my book, “Restricted Data: The History of Nuclear Secrecy in the United States.” In 1979, the Department of Energy and the Department of Justice attempted to prevent the publication of an issue of The Progressive magazine featuring an article by the anti-war activist Howard Morland that claimed to describe the then still-classified Teller-Ulam design for the hydrogen bomb. The decision to pursue prior restraint was not taken lightly: President Carter had signed off on it, and the Justice Department lawyers involved saw this as a post-Watergate/Pentagon Papers approach to thinking about what the boundaries of free speech might be—if not H-bomb secrets, then what else? 

Though they were successful in getting an initial injunction against publication, on appeal it became clear that this case was a lot less straightforward than they had assumed. The government had a difficult time establishing that drawings showing general weapons concepts could be directly connected to some future harm, and Morland and his attorneys were able to persuasively demonstrate that the information in his article had been pieced together from unclassified sleuthing. Furthermore, during the trial (which was split between unclassified sessions and classified, in camera ones), the government accidentally released considerable classified information. It became clear that the appeals judges were leaning toward considering that the “born secret” aspects of Restricted Data were possibly unconstitutional, but the government found an excuse to declare the matter moot before a verdict was rendered. Morland’s thesis about how the Teller-Ulam design works is now what one finds depicted by default when searching for information on H-bombs, having been “validated” by the attempted censorship.

There is much more that one can say about the case, but its major consequence was highlighting several key difficulties with prosecuting Restricted Data violations in the public domain. One is that while it might seem straightforward to claim that nuclear weapons design data can cause harm, legally it is a higher bar than one might assume, since the causal chain from “private person speculates publicly about how a nuclear weapon works” to whatever horrific consequences one might imagine could result is a very long one and requires many assumptions and leaps of logic. Another is that the prosecution of such violations can effectively “validate” and emphasize them, similar to what is today often called the “Streisand effect.” Finally, it emphasized that the constitutionality of the Restricted Data clause could not be taken for granted.

So can Trump declassify nuclear secrets?

One of the difficulties with all of these issues is that there is essentially no precedent in applying them to the president. Some precedent exists for  presidents trying to make certain things secret at will, or for the government trying to censor the publication of secrets, but really none when it comes to presidents asserting that they have declassified something secret and thus cannot be charged for mishandling it. It just hasn’t ever been an issue before that I can see. That means that it’s difficult to predict how the courts would rule on it, because on top of the First Amendment issues, you are also getting into deep questions about the power of the president—a nebulous area when talking about national security issues. Though, again, usually the issue has gone in the other direction—with presidents expanding secrecy, not reducing it. 

There have been instances of presidents declassifying National Defense Information by fiat. For example, Executive Order 12937 (1994), issued by Bill Clinton, declassified a list of documents in the National Archives by fiat, allowing information in them to be retained only if it was otherwise required to be for reasons other than national defense (for example, if they violate the privacy of a federal employee). The Project on Government Secrecy at the Federation of American Scientists says that this “unprecedented Order declassified approximately 45 million pages or 14% of the National Archives holdings of classified material, including classified holdings through the end of World War II, and an equal number dating into the 1970s.” This was part of a broad effort in the early Clinton administration to scale back Cold War secrecy, one that would ultimately leave him open to attack for being lax on security (notably by the Cox Report, which argued that lessened secrecy had allowed the People’s Republic of China to steal critical technical information, including on advanced thermonuclear weapons designs). There appears to have been no argument at the time, however, that the president could not declassify National Defense Information at will, that presidents can declassify documents, statements, or information at will, as they are the source from which the classification and declassification authority derives.  So there has definitely been some precedent for this happening in the past, but never in the way Trump or his lawyers have suggested it ought to work. There is an almost talismanic aspect to their description of it, with the president waving his hands over documents and declaring them to be declassified. Additionally, their description of how they treated the documents, including adding a “padlock” at the request of the FBI (offered up as proof that they were willing to work with the FBI) and keeping them in a safe, does not obviously indicate that they considered them declassified prior to them being seized.

If Trump’s purported declassification of these documents was to be taken seriously, one would expect that there would have to be a record of this somewhere, and that this would also mean that the underlying information in those documents would have to be declassified across the board: not just in those documents, but in any documents that contain them. Depending on what is in those documents, that would be a wide-ranging action with rippling effects as guidelines got updated accordingly. Clearly, it did not have that effect, or the FBI would not have seized them and declared them classified. In effect, Trump’s defense appears to be that the documents were secretly declassified.

For Restricted Data, the power of the president to declassify is even less clear. The updated version of the Atomic Energy Act that is currently on the books has detailed descriptions of how to remove information from the Restricted Data category. That process is initiated by the Department of Energy (as successor to the Atomic Energy Commission), not the president. The only explicit role the president has in this process is that if the Department of Energy and Department of Defense disagree on whether something should be declassified, the president acts as the tie-breaker. The president is given other explicit powers regarding Restricted Data, like the ability to direct the Department of Defense to share it with allied nations under certain circumstances (like planning for mutual defense, such as with NATO), but not declassification. The fact that the law does not explicitly give presidents the power to blanket declassify things, but does give them a role in declassification and other matters regarding Restricted Data, suggests that Congress’s intent was not to allow the president to declassify Restricted Data at will. 

Unlike National Defense Information, the procedures for identifying and declassifying Restricted Data are defined in statute, not in executive order. Whatever assumptions one might make about whether presidents need to follow their own executive orders (or the executive orders of previous presidents) or not thus get thrown out the window here as well. That doesn’t totally resolve the constitutional situation: Maybe you could try to argue that Congress doesn’t have the power to punish presidents for releasing Restricted Data, because that might interfere with their operations as commander-in-chief. Or, one could argue whether the Restricted Data clause is inherently unconstitutional, which as we’ve seen is not a new argument. Either way, it is a different issue at heart with Restricted Data than it is with National Defense Information. 

Of course, there may not be any Restricted Data in the documents from Mar-a-Lago. The category of “documents related to nuclear weapons” is vast, and not all such things would necessarily contain Restricted Data. For example, information about the North Korean, Iranian, or Israeli nuclear programs might be classified because of what they reveal about American intelligence sources or foreign intentions, not because they contain “nuclear secrets,” per se. And, of course, there may be no nuclear secrets of any sort in the seized documents at all: The tip might be a false one. 

One interesting consequence, though, of Trump’s current defense is that if he did truly declassify these documents, then they ought to be obtainable, at least in part, under the Freedom of Information Act. Are Trump, his lawyers, and his defenders suggesting that these documents should be publishable? That documents that the military, justice, and intelligence communities appear to think are important enough to keep controlled should just be released? (There are potential work-arounds here, like applying controlled-but-unclassified categorizations to them, but that’s another can of worms.) 

It does lead to the complicated question of “graymail” though: If Trump were to be prosecuted for these documents, and did claim that he had declassified them, and was successful in this claim, would this not mean the documents were truly declassified and thus could be compelled to release? Depending on the sensitivity of the documents, would that be enough to convince the government to avoid prosecution by itself? Without knowing more about the contents of the documents, this is all very idle speculation, but it does raise a lot of genuinely thorny and genuinely novel legal problems.

Alex Wellerstein is a historian of nuclear weapons at the Stevens Institute of Technology in Hoboken, New Jersey. He is the author of Restricted Data: The History of Nuclear Secrecy in the United States (University of Chicago Press, 2021).

Subscribe to Lawfare