Editor’s note: This article is part of the Just Security Symposium on Harold Hongju Koh’s “The National Security Constitution in the 21st Century”.

Harold Hongju Koh is one of the most accomplished international and U.S. foreign relations lawyers of his generation. Since publishing his first book, The National Security Constitution: After the Iran-Contra Affair (1990), he has served as the Department of State’s Legal Adviser, among other posts, and as Dean and then as Sterling Professor of International Law at the Yale Law School; in addition, he has litigated some of the most important international and foreign relations law cases of our time. It is only natural to expect that these experiences, and other intervening developments, might have shaped his thinking, and we are fortunate that he has reexamined his first book’s topics in view of that experience, resulting in The National Security Constitution in the 21st Century (2024). Koh’s new book is required reading for anyone interested in how U.S. foreign relations law is practiced in the United States and, more importantly, how those practices might be sensibly revised.

We focus here on one such practice, in which the president acts to withdraw the United States from international agreements. This issue was controversial when Koh first wrote and remains so—not least because, as Koh emphasizes (pp. 217-25), President Donald Trump withdrew from important agreements, such as the 1987 Intermediate-Range Nuclear Forces (INF) Treaty and the 1945 UNESCO Constitution, and threatened withdrawal from others, such as the 1949 North Atlantic Treaty that created NATO. The merits of this issue were not resolved by the Supreme Court’s 1979 decision in Goldwater v. Carter, and Koh expresses misgivings about claims of presidential unilateralism that have flourished in its wake, including aspects of the position adopted in the Restatement (Fourth) on the Foreign Relations Law of the United States (on which one of us served as a co-reporter).

To rectify matters, Koh urges adoption of a “‘mirror principle,’ whereby the degree of congressional approval needed to exit from an international agreement mirrors the degree of congressional approval needed to enter into that agreement in the first place” (p. 311). “[N]eeded” appears to mean either in theory or in practice, turning on “the subject matter of the agreement at issue and the degree of congressional approval involved in the entry into that agreement” (pp. 312-13; emphasis in original). The president would thus be disabled from acting unilaterally to withdraw from an agreement that the executive could not have made on its own (Koh gives the example of 1994 NAFTA) or, regardless of what might be constitutionally permissible, from an agreement that actually was made in the form of a congressional-executive agreement or treaty. Perhaps stretching the concept of “mirror” a bit, Koh maintains that even some international agreements that at the outset lacked formal congressional approval, like the 2015 Paris climate accord or the 2015 Iran nuclear deal, would by virtue of “legitimate congressional expectations” be removed from those the president could act unilaterally to exit (p. 313). Koh accepts that sometimes political exigencies may require a “quick divorce,” but favors only a presidential power to unilaterally suspend the agreement, thereafter securing whatever approval for termination a mirroring principle requires (pp. 221-22, 314-15).

As Koh recognizes, the U.S. Constitution’s text neither requires nor forecloses this “mirror” approach. The Constitution specifies how “treaties” are made (Article II, § 2), but says nothing about how other international agreements are made, and is entirely silent on the issue of exit. If one considers the constitutional drafting and ratification history, one finds an emphasis on the status of treaties as “supreme law of the land” (as Koh stresses, p. 222), but also finds a strong desire for the U.S. government to act in relation to treaties through the president with uniformity and dispatch, and finds no evidence of a “mirror” concept. Post-ratification practice does not embrace a “mirror approach.” We are aware of no case that ultimately reached such a conclusion (nor does Koh cite to one). The foundation of the modern controversy, President Jimmy Carter’s abrogation of the 1954 U.S.-Taiwan Mutual Defense Treaty, received real pushback in the form of litigation by some congressional members, but no formal action by the Senate or Congress as a whole in opposition. Something similar happened in 2002, when President George W. Bush withdrew from the 1972 Anti-Ballistic Missile (ABM) Treaty, which only resulted in an (equally unsuccessful) lawsuit by some members. In the Goldwater litigation, the D.C. Circuit—in a decision vacated, of course, by the Supreme Court—directly rejected mirroring, either as a description of historical practice or as an appropriate judicial rule. In approximately two dozen instances since then, the president has acted unilaterally to withdraw from international agreements of all types, without formal congressional opposition. To this might be added the Senate’s repeated consent to treaties (and Congress’ repeated consent to congressional-executive agreements) that contain articles specifically allowing for termination or withdrawal by a party (such as the Taiwan and ABM treaties), without any condition that the Senate (or Congress) be involved in such a decision. It is important to acknowledge, as Koh stresses (pp. 219-20), that congressional failure to act does not necessarily mean constitutional acquiescence, but none of the institutions involved seem committed to a “mirror” rule. It remains the case, just as the D.C. Circuit Goldwater majority said, that “in no situation has a treaty been continued in force over the opposition of the President.”

As such, we are hesitant about the strong claim that “U.S. constitutional jurisprudence has long acknowledged that withdrawing from international agreements should be dictated by a mirror principle” (p. 312). The last-in-time rule, holding that conflicts between a statute and a treaty should be resolved in favor the later-in-time instrument, which Koh invokes to support his mirror principle (p. 312), itself suggests a warp in any mirror. Under that rule, the Senate and president may, through a treaty, supersede a statute previously enacted by Congress as a whole; the statute need not be repealed by Congress. Likewise, a congressional majority (or supermajority, if overriding a veto) may supersede a treaty previously approved by two-thirds of the Senate; exit from the treaty need not be consented to by a Senate supermajority. Given the perceived equivalence for this purpose of treaties and statutes, Koh is correct (p. 312) that it is somewhat anomalous to allow the president to repeal the “law of the land” when it is in the form of a treaty but not in the form of a statute. But the treaty process itself is anomalous as compared with the statutory process. And, in any event, the last-in-time rule simply does not address the question of when, or by whom, a valid source of law may be terminated (for example, the rule does not preclude a court from striking down a statute). We tend to agree with Koh (p. 312) and others, like Kristen Eichensehr, that presidential authority over appointments is not precisely comparable to the issue of terminating international agreements, but presidential authority in relation to treaties and other agreements is distinct from the presidential role in legislation as well.

Should we embrace a mirroring principle nonetheless? Koh’s argument is addressed not just to the courts but also to executive branch officials, who can chart such a course even if they strenuously resist its enforcement by the judiciary. There may be much greater consensus here. First, everyone apparently agrees that the executive branch can unilaterally terminate an agreement when Article II powers were the sole, and a constitutionally sufficient, basis for entering into it.

Second, although Koh does not directly address the issue, there may be consensus that the president can unilaterally terminate an agreement if he or she determines that the agreement has ceased to be binding on the United States as a matter of international law, such as due to another state’s conduct (e.g., material breach), due to circumstances such as force majeure or impossibility, due to the effects of armed conflict, or due to a superseding later-in-time treaty or rule of customary international law. Likewise, there may be consensus that the president can unilaterally terminate an agreement that is no longer consistent with U.S. law, such as due to the adoption of a subsequent statute that is flatly inconsistent with the agreement. In all such instances, unilateral presidential termination appears fully consistent with the Constitution, including the president’s obligation to take care that the law be faithfully executed.

Beyond these circumstances, situations of exit are not best resolved through a single principle but, rather, a more fact-sensitive approach. The issue has not been joined, but Koh and the Restatement (Fourth) both suggest that Congress or the Senate might limit presidential termination, at least when they indicate as much in authorizing the agreement, if not through a later statute (as was done in relation to potential withdrawal from the North Atlantic Treaty). Possibly, Congress or the Senate might register such limitations implicitly, though it is highly likely that there would be debates about whether that has happened. If the executive ever acted in disregard of such limits, that would appear to establish a conflict falling into Justice Robert Jackson’s third Youngstown category—in which the president acts in a matter incompatible with the express or implied congressional will—which is quite unfriendly to presidential authority. Further, it would likely create problems for the executive in securing future consent from Congress for new international agreements. Conversely, Congress or the Senate might authorize unilateral presidential termination, in which case such termination should not require mirroring at the point of exit.

The most common situation is when Congress or the Senate is silent when approving an agreement. Often the agreement will contain a termination clause, as was invoked by Carter when terminating the Taiwan treaty and debated in the Goldwater litigation. If viewed as an explicit or implicit delegation to the president of a power to terminate unilaterally, then this situation would fall into Jackson’s second category (in which the president acts absent either a congressional grant or a denial of authority, permitting some capacity to act independently) at worst or his first category (in which the president acts with express or implied congressional authorization, putting executive authority at its maximum) at best. Most recent presidential terminations (the ABM treaty, the Paris climate accord, the INF treaty) have in fact been effectuated by the president pursuant to a termination clause in the agreement. Some agreements do not contain a termination clause—the U.N. Charter or the International Covenant on Civil and Political Rights being examples. Here the president cannot rely on any explicit or implicit delegation based on a termination clause, but perhaps can do so based on the international law of treaties, of which Congress and the Senate are aware. Under that law, there is essentially a presumption (reflected in Article 56 of the Vienna Convention on the Law of Treaties) that a party cannot withdraw unless it can be established that the parties intended to admit the possibility of such withdrawal or it can be implied from the nature of the treaty. That presumption might place unilateral presidential termination in Jackson’s third category, while successful invocation by the president of the exceptions might place it in either the first or second categories.

For all cases in which the president is acting consistent with a treaty termination clause or with background rules found in the law of treaties, we think it is best to avoid describing such termination as “agreement breaking” (Koh’s term), since the proposition in question is how the United States, as a matter of U.S. law, may exit from an agreement when doing so is in full compliance with international law.

As a practical matter, the mirroring principle would shift focus from termination proper to whether the underlying agreement formally involved congressional or Senate approval, or generated congressional “expectations” of some kind. If the ideal outcome is greater inter-branch cooperation in making international agreements, we are wary of a principle that tends to discourage such cooperation at the outset—in effect, telling the executive branch that if it secures legislative authorization or other support to bolster the legal or political grounds for entering into an agreement, the executive will be encumbering its ability to terminate that agreement. At present, there is some beneficial “play in the joints” (p. 331) for agreement-making. Whether an agreement falls outside the president’s unilateral capacity, and instead involves congressional subject-matter authority or requires Senate approval as an Article II treaty, is sometimes unresolved when the United States enters into an agreement—not least because of disagreement about whether there is implicit legislative authorization of some kind. A mirroring principle might reduce such flexibility, perhaps ultimately encouraging the president to make increasingly aggressive claims that he or she can go it alone when making international agreements. Moreover, if that issue has to be resolved whenever the president decides to terminate unilaterally, even when Congress raises no objection to such termination decision, it might lead to distracting doubts in the United States and abroad about whether the agreement was lawful in the first place.

On balance, a mirroring concept serves better as part of an aspiration for greater inter-branch cooperation in making and unmaking international agreements, rather than as a legally-enforceable rule. Prevailing law and practice is far more differentiated and fact-sensitive on this issue—a conclusion we think is quite consistent with Koh’s overall, compelling objective of avoiding a rigid rule whereby the president can always terminate agreements unilaterally.

IMAGE: The Harry S. Truman Federal Building, headquarters of the U.S. Department of State, is pictured on October 08, 2024 in Washington, DC. (Photo by Kevin Dietsch/Getty Images)