Engaging China and Russia on Arms Control: An Interview With U.S. Assistant Secretary of State Mallory Stewart

May 2024

During an era when nuclear weapons threats are growing, Assistant Secretary of State Mallory Stewart is among the top U.S. officials working to increase stability, prevent conflict, and preserve and advance effective arms control and disarmament measures. The office that she leads, recently renamed the Bureau of Arms Control, Deterrence, and Stability, is charged with implementing and ensuring compliance with existing arms control agreements and negotiating new ones. A current focus is pressuring Russia to return to compliance with and identify a follow-on agreement to the New Strategic Arms Reduction Treaty (New START), which expires in 2026 and is the last remaining treaty limiting the two largest nuclear arsenals. So far, Russia has refused to engage until the United States withdraws its support for Ukraine, where Russia has waged an all-out war since 2022. Stewart’s bureau also is working to jump-start nuclear risk reduction consultations with China, which has a fraction of the nuclear weapons possessed by Russia and the United States but has greatly accelerated its nuclear weapons program. Working bilaterally and in multilateral forums, the bureau also has a goal of promoting responsible behavior with emerging military technologies, including artificial intelligence. Daryl G. Kimball, executive director of the Arms Control Association, and Carol Giacomo, editor of Arms Control Today, explored these issues in an interview with Stewart on April 4. This transcript has been edited for clarity and length. 

ARMS CONTROL TODAY: Russia has rejected the U.S. proposal for talks to identify an arms control framework after New START expires in February 2026. What is realistically possible before the next administration takes office? 

Assistant Secretary Mallory Stewart: Certainly, we’ll continue to reinforce what National Security Adviser Jake Sullivan said last June, which is that we’re willing to engage with Russia and China without preconditions. But he was very clear to point out that engagement without preconditions doesn’t mean engagement without accountability. We have made good faith efforts toward this end, including in our September 2023 diplomatic note to the Russians suggesting possible paths forward toward this post-New START era. The Russians rejected it three months later, tying it specifically to U.S. support for Ukraine in its defense against Russia’s illegal and continuing invasion. 

This refusal to engage on arms control is deeply irresponsible of Russia, which has a commitment to the international community, just as the United States does, to not only fulfill its Article VI obligations of the nuclear Nonproliferation Treaty (NPT) but also pursue risk reduction, global stability, strategic stability, and all of the requirements that the international community expects of us as the two largest nuclear-weapon states under the NPT. We’ve tried to make clear to Russia that these are responsibilities that we share, regardless of Putin’s ambitions to take over Ukraine. We understand that we can’t force them into an arms control discussion, but what we can do is try to work with the rest of the multilateral and international community to make the case for pursuing risk reduction measures and building an increased appreciation for why it is in Russia’s interest to engage in arms control conversations.

We’ve been trying to take a twofold approach in that endeavor. First, we are working to combat the disinformation that Russia persists in purveying and, unfortunately, that some countries agree with, that somehow the invasion of Ukraine by Russia was justified. That narrative is factually incorrect and doesn’t reflect how we were very actively engaged with Russia prior to the invasion of Ukraine. We addressed every element of their proposed treaties to the United States and NATO and explained why it wasn’t realistic to suggest the United States was going to be able to turn back time to 1997 or to take other actions with respect to other sovereign governments to prevent this invasion. It’s unfortunate that many governments blame NATO and the West for what Russia was intending to do when the U.S. government specifically engaged with Russia to try to prevent this invasion. We even warned Russia in that conversation that if they invaded Ukraine, NATO presumably would expand because NATO is there to protect and defend against illegal aggression. 

We’re also correcting the disinformation with respect to Russia’s illegal suspension of New START. The Russians prevented the bilateral consultative commission from meeting. That commission is specifically included in New START to address the concerns of either party regarding verification and implementation. We had been making progress prior to the COVID-19 pandemic on addressing concerns of both sides. Russia then saw it as in its interest to try to leverage our desire and our responsibility to continue New START as some sort of blackmail to try to get us and our partners and allies to stop supporting Ukraine. There was no justification for their suspension of New START, and it is deeply irresponsible.

Second, the plan going forward is to continue to work with the multilateral community in order to enhance transparency, demonstrate responsible behaviors, reduce the chance of misunderstanding and miscalculations, and prevent unintentional escalation. Russia must uphold its international obligations and engage in risk reduction and arms control-oriented discussions. We will also continue with the P5 [the five nuclear-weapon states recognized by the NPT] expert discussions on risk reduction, hopefully getting some substance out of all P5 member states toward discussing nuclear doctrine, clarifying misunderstandings and avoiding miscalculation, and discussing risk reduction possibilities. 

ACT: What are your concerns if February 2026 comes along, New START expires, and there’s no agreement on a follow-on framework? 

Stewart: The concerns are what we face right now, only worse. Our intelligence community was able to confirm its belief that Russia did not, as of the end of last year, exceed the caps set by New START. Russia has said it will abide by New START caps, and we feel that we’ll be able to respond in a reciprocal way. But without New START, we’re not able to perform the verification procedures and feel as confident in our compliance assessments. 

So, the concern, come 2026, is that we will have no guardrails in place, thus risking misunderstanding and miscalculation. Quite frankly, the biggest risk is an unbridled, unconstrained nuclear arms race, which no one wants, no one needs, and the international community should be outraged that we’re now stumbling toward that eventuality. 

ACT: You mentioned the P5 process. What is possible at the NPT preparatory committee meeting in July given that the P5 group is just meeting at the experts level and Russia is the current chair?

Stewart: We have to engage and strive toward risk reduction. Russia may not be inclined to agree with us on that, but we will keep trying because in those P5 conversations, we explain why it’s in their interest to prevent misunderstanding and miscalculation and why risk reduction, which they fundamentally have understood since the Cold War as in their interest, is still in their interest and is something that the P5 owes the international community. We should demonstrate that we can all be responsible actors. 

Hopefully, we also can make some progress toward sharing nuclear policies to at least correct misunderstandings. This was the hope we had in engaging with China as well. The P5 is a good process to say we can address your questions, please address our questions, let’s move forward on that basis. So, the fact that the P5 is still going forward under Russia’s leadership is a good sign, but we are pushing for more substantive results from these conversations. The international community expects that of us. 

ACT: In February, the Biden administration warned that Russia is believed to be pursuing an anti-satellite (ASAT) program that would violate the Outer Space Treaty and could involve a nuclear explosive device. On March 18, the United States and Japan announced that they were pursuing a UN Security Council resolution to reinforce the treaty. What else is the United States doing to build support for the treaty? 

Stewart: The Security Council process is really important. In that context, we’ve provided a lot of information as to why this resolution is important now. There are questions from Security Council members, and we’re addressing them. As much as possible, we’re talking to other governments about our concerns and about how irresponsible, destabilizing, and really catastrophic a placement of a weapon in orbit would be for the future of the spacefaring generation, given that going forward every single country is going to rely on space more and more. We are sharing as much information as we can. Even if Russia is denying the weapon’s existence, it does make sense to go forward in the multilateral arena and reconfirm commitments to the Outer Space Treaty. 

ACT: Related to that, what is the status of the U.S.-led initiative on prohibiting ground-launch kinetic intercept tests? 

Stewart: We have large, broad consensus. Some 155 countries signed on to the resolution in the UN General Assembly, but we have 37 countries that also have made a national statement or passed some type of domestic implementation legislation. We are continuing to push that forward so that more countries can join the United States in turning this high-level, multilateral political commitment into a domestically legally binding commitment. 

ACT: What end state does the United States seek to attain with this initiative? 

Stewart: Ultimately, if we can get as many countries as possible to enforce this legally in their domestic capacity, it can make it easier when we talk with the broad multilateral community about what is possible in the legally binding arena. 

In all of our efforts on risk reduction, especially with respect to emerging technologies with strategic effects, there’s been a misnomer that somehow the United States is pushing back on legally binding instruments. We’re not. We’re trying to lay a general understanding as to what is possible, verifiable, and achievable. We hope to come to consensus on what should be prohibited, as a means to define either responsible or irresponsible behavior. Ultimately, many countries want something that is legally binding. We are sympathetic, but we need to be in a place where saying something is verifiable or accountable through legal means reflects an ability to ensure compliance. 

We’ve seen where countries propose treaties, such as the prevention of placement of weapons in outer space treaty that Russia and China have pushed for many years, without even an accepted common definition of what a “weapon” in space is. There is an inconsistent appreciation for what the terminology of weapons in space means right now, and there is no way to verify compliance or to even verify steps taken that are consistent with that treaty. They haven’t come to a common understanding of the definitions of these principles, and if the multilateral community signed on to a treaty without a common understanding of what the terms mean, it is unenforceable. 

So, the U.S. position is, Why not start from developing a general understanding of the things that we agree are harmful to international security, such as direct-ascent destructive ASAT testing, and build from that? The good thing about this effort is that it is to a large degree verifiable. You can see a direct-ascent destructive ASAT capacity come from earth and create debris in space. So, by trying to build on those concepts, ultimately, in a better world with more security, stability, understanding, and agreement, we can get to a legally binding instrument where we have the parameters of common terminology, an agreement on definitions, and some way to ensure that that treaty is being complied with. 

ACT: In November, you met your Chinese counterpart for arms control talks. What is the status of the U.S. effort to engage China on risk reduction and arms control? What obstacles are you seeing, and what specific outcomes are you trying to achieve? 

Stewart: The Biden-Harris administration has talked a lot about our desire for preventing unintentional escalation, misunderstanding, and miscalculation—the classic desires and end goals that we strive for in the arms control arena. We made very clear we would like to engage with [China] again, but we need to engage in a context in which we can both share concerns and take steps to substantively address those concerns. We also need participation in this conversation that reflects the best expertise to address the concerns of both sides. It’s why the United States was very specific in bringing an interagency group to our November 6 meeting. The Chinese side did have, to a certain degree, different agencies involved on their side; but going forward, we really need to have substantive engagement and ideas that we can put into place to reduce the potential for misunderstanding and demonstrate that we’re moving toward our obligations in the international community. 

A large part of the obstacles we are facing is fundamental disagreement over basic concepts. From the U.S. perspective, we see transparency as stabilizing. We are very forward leaning to be as transparent as possible, especially with those governments with which there may be a decrease in trust. In the P5 expert conversations, we try to demonstrate clarity and answer questions with respect to our nuclear policies. We see this as positive for U.S. security and also for our relationships with other governments.

We’ve heard the recent statements from [China] that there is desire in Beijing to engage in a conversation on no first use, and I have a lot of questions about that. When we talked to the Chinese, we wanted, and we still very much hope, to engage in risk reduction, and part of that included proposals that would expand and enhance transparency. The reason we need to do that is to build toward what China has acknowledged is a trust deficit. We need to build toward them understanding our policies and us understanding what the international community sees as potentially not explained in their policies with respect to their massive nuclear buildup. 

Our questions to China are, Help us understand how your nuclear buildup is consistent with your no-first-use policy, help us understand how what we’re all watching, which is this launch-on-warning capacity, is consistent with China’s statement from 2019 that they don’t need such a capacity and that it would be inconsistent with their no-first-use policy. We don’t want to assume the worst, but again, in this sort of trust deficit environment, the assumption will not be positive, and so that’s why we need to engage on these questions and have these conversations. 

ACT: You noted the Chinese talk of a mutual no-first-use agreement. They seem to describe that as a risk reduction approach. Whether it is or it isn’t, is the United States willing to substantively engage on that idea? Your description of their response is confusing. Isn’t that a signal that China is interested in talking about risk reduction options, even if their approach differs from what the United States is proposing? 

Stewart: I would love to ask them all of these questions and engage on this. I would love to ask them to explain how you can have a no-first-use policy if you just declare it and when everything you’re doing or developing seems inconsistent. If they want to engage in a conversation of the many questions raised by their no-first-use proposal, we would engage.

But in that same conversation, we should talk about other risk reduction proposals that we’ve put on the table that are as minimal as missile test launch notifications. This would be similar to the launch notification agreement that China already has with Russia. Those ideas are on the table, but so far have not been taken up by China. 

ACT: To clarify, you held a first arms control meeting with China. You had questions. They had proposals. Were the Chinese just not being forthcoming? 

Stewart: To be clear, there weren’t any proposals that they gave to us. But we had questions, and they had questions, and it was the first round in several years. I thought it was a constructive step to demonstrate that our questions are legitimized by openly available information. They had questions of us, and I encouraged those questions. The more we can answer, the better. 

I think that the reason that there is a desire now for more substantive engagement is that there were many questions we had that couldn’t be answered by China…. The reason we want to take steps to move toward risk prevention or reduction before we can even talk about something in the legally binding arena is because we have to develop that what we say and what they say matches what we’re both doing. 

ACT: Linda Thomas-Greenfield, the U.S. ambassador to the United Nations, said that the Chinese are refusing to engage substantively. Have you asked them to schedule another arms control meeting? Was there any discussion of these issues when U.S. President Joe Biden and Chinese President Xi Jinping spoke recently on the phone? 

Stewart: The specifics on engagement were left to sort out for a future discussion, but in the official Biden-Xi readout, there was a discussion of additional conversations on our issues. At the highest level, there is support for a continued conversation. We have made clear that we are willing to engage, and we would like to engage…. But we also need these engagements to be substantive and to reflect that we both fully appreciate the ability to ask questions but also to answer questions. 

ACT: Did the Chinese say, “We don’t want to put proposals on the table,” or did they just not answer yours? What exactly is the state of play?

Stewart: We’ve put some positive proposals on the table, and we also welcome their positive substantive proposals. This is why I think we should be willing to ask questions and engage because if they have a risk reduction proposal that would require first the development of a level of trust, appreciation, or understanding, then let’s figure out how we develop that trust, so we can get to something that would work in the risk reduction arena…. But I haven’t seen any actual [Chinese] proposals aside from hearing of this no-first-use proposal.

ACT: The Conference on Disarmament (CD) in Geneva remains deadlocked on negotiations over a fissile material cutoff treaty (FMCT) and other issues. At the UN Security Council meeting on March 18, the United States and Japan proposed a Friends of FMCT group to move the proposed treaty forward. How could that change the dynamics? 

Stewart: The United States is very proud to support the Japanese initiative on the Friends of FMCT. The initiative reflects strong support for an FMCT and the need to address it because of the challenges we face. From the U.S. perspective, we have tried to demonstrate flexibility with respect to some of the deadlocks in the process at the CD. For example, we are open to discussing negative security assurances, which is something that a lot of countries deem as very important. In return, we have asked that those countries be open to the initiation of FMCT discussions. 

I certainly can’t say that we’re overly optimistic, but we are trying to engage new partners and new conversations and show flexibility on what have traditionally been hold-ups for the United States. We want to demonstrate that we are willing to engage in these conversations, raise our concerns, hear other states’ concerns, and try to move beyond the politics, and that’s been one of the biggest problems. I hope the Friends of FMCT will be a positive opportunity for broader conversations, with maybe even new voices, not just the traditional voices that have deadlocked so adamantly in the CD. We can only try our best, and we will continue to do so. 

ACT: In 2018 the Trump administration declared that it will not seek Senate ratification of the Comprehensive Test Ban Treaty (CTBT). The Biden administration has made clear that it has no need or any plans to test and that it is committed to working to achieve CTBT entry into force. But will it seek CTBT ratification? 

Stewart: The Biden-Harris administration supports the CTBT and is working to bring it into force. However, the environment is getting even worse internationally with Russia withdrawing its ratification from the CTBT. What we are happy about is that the CTBT International Monitoring System (IMS) is nearing completion, thanks in large part to a lot of resources and expertise from the U.S. government. We need to build appreciation for the IMS capacity to detect explosions and for a concept that the Biden-Harris administration deeply appreciates, which is that the CTBT in force is better for the entire global community. 

We are encouraging China to continue to push forward on the monitoring stations in its own country, as well as encouraging the international community to explain to Russia that its withdrawal was not the right approach and not a responsible action. On our part, there is no U.S. interest, desire, willingness, or support for testing. It’s not necessary, it’s not helpful, it’s deeply destabilizing, and it’s not something the U.S. government will do. 

ACT: In order to improve its credibility, isn’t it necessary for the administration to state unequivocally that it still intends to ask the Senate to provide its advice and consent for ratification of the CTBT because the United States is one of the 44 countries that must ratify for the treaty to enter into force?

Stewart: This administration has been very clear. We support the CTBT, but we can’t ratify the treaty just as a sole executive decision. So, we need to work with our congressional partners and within our political system to foster better understanding of why this is a valuable treaty and how it protects our national security. If we just lean in right now without any further change, we’re going to have the same stalemate that we’ve had on this treaty for years on end. 

ACT: The State Department has been developing support around a political declaration on military uses of artificial intelligence (AI). Other people are pushing a legally binding agreement. Are these approaches compatible? Why not work for a legally binding format? 

Stewart: I actually see them all as consistent. We supported the Austrian resolution at the United Nations. We support the conversation in the UN group of governmental experts going forward on lethal autonomous weapons. The Austrian initiative, with respect to lethal autonomous weapons, says we should look at this issue more broadly. Our political declaration addresses a unique and different context, but is entirely consistent in the sense that we are building a common understanding of the potential risks of AI across the full range of military applications. 

Whether in human resources decisions or weapons systems, AI could carry risks across the board if you don’t appreciate in advance that there are advantages but also potential challenges it could bring into the system. The political declaration is one unique context for concrete development of ideas and actions for governments on the larger issue. We are now up to 54 countries endorsing the political declaration and 14 observing states attended the recent plenary meeting, and we’re hoping to build that out. 

Austria is one of our working group chairs, and we’re really excited about that. The United States is also chairing a working group with Bahrain on assurance, covering technical measures to establish justified confidence in the reliability of AI capabilities, and Portugal and Canada are chairing a working group on accountability measures. These three working groups will be used to push forward on concrete steps to develop greater understanding, appreciation, and actions we can take with those countries to minimize risk. 

We hope more countries join us. We look forward to making progress in understanding how we can ensure that AI operators actually have some ability to understand context, understand risk, and implement training at a knowledge level that prevents those risks. 

ACT: So, you see a place for a legally binding agreement?

Stewart: Eventually, yes, if we can come to a consensus on the best path forward. The challenge, of course, is that these emerging technologies evolve so quickly that if you lock that into a legally binding agreement without incorporating all the possible changes and iterations of how AI can manifest, you may lock yourself into a situation that evolves next week.

To compare, we’ve seen the need to update the schedules on chemical weapons, and it’s great that the Chemical Weapons Convention allows for states to do that. But we don’t even have a standard definition of what AI is with respect to its use. Is it machine learning, is it intelligent learning, is it something in between? Is AI a defined concept that everyone agrees on? Not right now, and we need to work to understand that before we can embed it in a legally binding treaty. 

ACT: What kind of things is the political declaration intended to prevent? It was reported recently that Israel has been utilizing an AI-enabled system called Lavender to identify targets in Gaza that is not being reviewed carefully by humans to determine whether there are errors in the system, thus possibly creating additional casualties. The political declaration contains principles to avoid the risk of automation bias. I’m not asking you to judge whether the reporting on the Lavender system is accurate, but isn’t this kind of automation bias what the political declaration on AI says should be avoided?

Stewart: Automation bias is an example of a situation that we would hope countries can take steps to prevent in their use of AI. I cannot speak specifically to reports regarding the Lavender system, but from what we’ve discussed already and the principles in the political declaration, we should be able to understand the appropriate usage of AI and the risks that AI brings in and take steps to prevent those risks. 

One basic concept that’s in the political declaration is related to a scenario where a military might design AI for a certain use, but then through the course of operation, switch it to a different deployment context without changing anything about the AI. This is a risky situation because you’re now putting AI into an entirely different environment that it was not designed for or you’re not adapting the AI system to parameters that can exacerbate some of the challenges that AI and autonomous functions bring to the situation. These are the kind of things that we have been discussing, and we look forward to figuring out how we take steps to prevent.