The Hon. Christopher A. Ford
New Paradigms Forum -- International Security Policy Since 2009
Blog Layout
Challenges of Weapon Technology Limitation, 1874-2021
Dr. Christopher Ford • Apr 16, 2021
Below follow the remarks Dr. Ford prepared for his speech to the U.S. Air Force Futures Office on April 15, 2021.
I am especially pleased to talk with your team today, because I am both personally and professionally interested in the ways in which emerging technology issues intersect with U.S. national security and foreign policy, American competitiveness in the face of great-power challenges, and the future of warfighting. Any views I express here will be entirely my own, of course, and shouldn’t be taken necessarily to represent those of anyone else in or out of government. Nevertheless, what I’d like to do today is say a few words – from my perspective as having until recently been the seniormost U.S. official concerned with arms control and nonproliferation – about how the broader policy community relates to novel and potentially disruptive developments in the arena of military technology.
I. Longstanding Efforts to Regulate Disruptive Technologies
These aren’t precisely new questions, of course. The sometimes bewildering pace of advancing military technology in the industrial age – and now the information age – has for a long time presented policy community leaders with repeated challenges about whether, when, and how it might be possible to regulate such technologies, and their use.
A. Early Efforts
There has certainly not always been general agreement that such regulation is even possible, or how effective it can be relied upon to be. Nevertheless, at least some European leaders began to see a need for new rules to govern the conduct and tools of war beginning in the mid-19th Century. In reaction to dawning realization of the sort of carnage that industrial-age warfare could produce – upon which people began to focus at least as early as 1859, when Switzerland’s Henry Durant was inspired to create the International Red Cross upon viewing the horrid spectacle of some 40,000 soldiers lying dead or dying upon the field after the battle of Solferino between Austria and the then-Franco-Sardinian Alliance – interest grew in trying to devise negotiated means of mitigating such ferocity.
The United States deserves credit for helping take seminal early steps toward a new system of regulating warfighting, for it was the Union Army during our Civil War that drew up the first modern codification of the laws of war. This was President Abraham Lincoln’s General Order No. 100 of 1863 – his “Instructions for the Government of Armies of the United States in the Field,” today remembered as the “Lieber Code” after its author, the German-American lawyer and professor of history and political science Francis Lieber. That remarkable effort did not deal with issues of military technology – being more focused upon military conduct
– but it still forms the foundation of present-day legal guidance for the U.S. armed forces, and was used as a model when multiple other powers produced their own codifications, greatly advancing the development of International Humanitarian Law (a.k.a. the Law of Armed Conflict, or “LOAC”).
On the initiative of Tsar Alexander II of Russia, a conference was held in Brussels in 1874 to draw up an International Declaration on the Laws and Customs of War, though the document it produced failed to win much support from governments of the era. Another effort was held in the form of the Hague Peace Conferences of 1899 and 1907, which attempted to spell out rules for conduct during war, accountability for war crimes, and the peaceful settlement of disputes. By the time of the second conference in 1907, the participants tried – unsuccessfully – to pivot into arms control limits on naval assets in response to the worsening naval arms race then underway.
For present purposes, however, it’s worth noting the degree to which the Hague conferences also attempted to craft restrictions on certain novel and potentially disruptive military technologies of the period. In particular, the 1899 conference adopted declarations against discharging or dropping explosives from balloons, the use of projectiles to spread poisonous gases, and bullets that expand or flatten inside the human body (including soft-tipped, so-called “dum dum” bullets). These were not exactly general
provisions, for these rules were only to apply between signatory powers, but they do represent an early effort to regulate military technology applications.
Arguably illustrating the challenges of arms limitation diplomacy in the face of accelerating great power competition an arms racing, of course, all this well-intentioned effort wasn’t up to the task of keeping the major powers from bleeding each other all but dry on an unprecedented scale beginning in 1914. Nor did it seem to do much to mitigate the horrors of a war that slaughtered millions and saw the first major use of chemical weapons, aerial bombardment, and unrestricted submarine warfare.
B. The Interwar Period
Spurred on by grim memories of the Great War, however, the major powers tried again in the 1920s and into the 1930s, including in a naval disarmament conference in 1921-22, the Geneva Gas Protocol of 1925
banning the use (although not the possession) of poison gas or bacteriological weaponry in war, a League of Nations-sponsored Preparatory Commission intended to lay the groundwork for general disarmament, the Kellogg-Briand Pact of 1928
that in effect purported to solve all these problems at once by forswearing war as a means of settling disputes. It also included an abortive U.S.-sponsored effort to call a naval conference, and another naval conference in London in 1930.
Of these, the Washington Naval Conference of 1921-22 is best known today for having set specific ratios between the tonnage of the major powers’ naval fleets and declared a halt to capital ship construction, and even limited the novel technology of submarines by codifying tonnage parity between the United States, Britain, and Japan. It also reached agreement, however, on stopping the fortification of Pacific Islands, and on multi-power consultations in the Far East, and on respect for China’s territorial integrity – as well as providing the locus for negotiations that led to bilateral agreements on such things as Japan’s return of the province of Shantung to China and withdrawal from Siberia.
With the London Naval Treaty of 1930, the United States and Great Britain were able to reach a compromise on language with regard to the classes of “cruiser” that had made agreement impossible before, and tonnage limits were thereupon imposed on the basis of fleet-by-fleet ratios. Additional submarine restrictions were established too, both in terms of tonnage and in terms of wartime conduct vis-à-vis commercial surface vessels. Destroyer tonnage was also limited. After London, another naval conference was held in 1932, and an additional naval treaty negotiated in 1936.
One of the things that kept this final phase of interwar arms control from being viewed as more of a success, however, was that Japan withdrew from the London conference of 1936 – with predictable results in undermining hopes for enduring compliance by other countries even with existing limits, as Tokyo built its way up toward a broader Pacific War – and that Mussolini’s Italy refused to sign, it being busy invading Ethiopia at the time. As for Nazi Germany, it had withdrawn from the League of Nations pursuant to a referendum held in 1933 and was also busily rearming.
As a consequence, the final stages of the interwar arms control process ended up consisting of limitation agreements negotiated only among powers that didn’t
by that point particularly fear each other, yet which did
all face an unconstrained and growing menace from what would in a few years’ time be the belligerent Axis Powers of the Second World War. The results are well known.
For purposes of understanding the challenges regulating military technology, however, perhaps the most interesting of these interwar efforts were two that clearly didn’t
work, even on their own terms: (1) the technical discussions associated with the League of Nations’ hopes to convene a conference on full-scope arms limitation or disarmament; and (2) the effort by U.S. President Calvin Coolidge to convene a second naval conference in Washington.
The long and intricate League-sponsored talks in 1926-27 did not end up producing agreement, however, as the parties were not able to identify a way of defining the problem, measuring military power, setting appropriate limits, and ensuring compliance amidst the interlocking complexities created by having so many very differently-situated parties at the table in an era of technological change. Those debates, however, provide a fascinating window into the breadth, ambition, intricacy, and intractability of the interwar disarmament problem.
On air disarmament, for instance, it was recognized that aircraft were important new and potentially decisive tools of war, and earnest efforts were made to figure out how they might be regulated as part of a broader scheme for limitations across the full spectrum of military power. To this end, debates flared at the League’s Preparatory Commission over whether to measure levels of military airpower on the basis of total engine horsepower, aggregate liftable cargo or weapons payload, or numbers of air service personnel. (Each method had its champions, who naturally promoted the metric that most advantaged – or at least that least disadvantaged – their own air services.)
Considerable controversy also arose over what was, in effect, a problem of dual-use technology. Specifically, it was a key early question whether and how to regulate civil
aviation, inasmuch as civilian aircraft could be repurposed in various ways for military use, creating potential workarounds for aerial disarmament, especially for countries with advanced civil aviation industries. There were also disagreements over how to deal with military aircraft based on land versus based on aircraft carriers, especially as those vessels had been subject to some limitations at the first Washington Conference.
Some participants in the Preparatory Commission debates hoped to be able to limit aviation-related matériel
as well, but it was also recognized that aeronautical science was progressing rapidly, and that what sorts of matériel
were actually relevant might change considerably in the future. Fast-advancing technology has always challenged arms control regimes!
Remarkably, these various discussions also struggled with issues such as whether overall military power could be measured, whether one could somehow distinguish between “offensive” and “defensive” armaments, how military budgets
might be compared and limited, and how to compare and limit land forces between powers that all had differently-sized standing armies, different personnel mobilization systems and capacities, and industrial bases of varying size and sophistication. Perhaps not surprisingly, the delegates did not end up producing enough agreement on such technical issues to support the general League disarmament conference that had been anticipated.
President Coolidge’s initiative for a second naval conference in 1927 also fell apart, but it too provides an interesting window into the challenges of such negotiation. The principal problems were between the United States and Britain over how to handle classes of naval “cruisers.” Both powers felt cruisers to be vital to their far-flung geographic interests, but these ships had not been much regulated at the Washington Conference, and an arms race was then underway. The details are sometimes arcane, but in effect, the two countries’ positions – driven by strong perceptions of a need for cruisers but also by different views of what ships it was safe for the other
party to have – never allowed London and Washington to quite agree on the gun caliber permissible for smaller cruisers, and the number and tonnage of larger vessels. (Ironically, as the historian Dick Richardson has observed, these irreconcilable U.S. and British negotiating positions in a sense both
derived in part from assessments in Washington and London of what they each needed in the face of a growing naval threat from Imperial Japan in the Far East. Threats from a third party, therefore, helped ensure bilateral negotiating failure.)
As history records, these various interwar efforts had rather mixed success – the checkered nature of which can be seen in the very fact that we now remember that era as the “interwar” period. To be sure, the Washington Naval Conference, in particular, succeeded for a time in regulating naval arms racing in at least some types of vessel. It also managed to kick the proverbial “can” not unhelpfully down the road in the Far East. As time went by, however, efforts to advance the agenda further stalled, after which the global military system went into retrograde motion.
Ultimately, these efforts proved unequal to the all but impossible task of handling the eventual coincidence of at least five major challenges: (1) America’s rising naval power vis-à-vis Britain; (2) Germany’s desire to rearm in the face of French neuralgia after the notably vindictive Versailles Peace of 1919; (3) Italy’s growing militarism under Benito Mussolini; (4) Japan’s increasingly aggressive imperial hegemonism in Asia, especially China; (5) the rise to power in Germany of the aggressive and expansionist Nazi regime; and (6) the impossibility of negotiating the ultimate objective of general
disarmament (as opposed merely to arms limitation), at least in this context. It would not be quite right to say that the interwar arms control and disarmament experiment was an abject failure, for a degree of arms restraint – and even of disarmament with regard to at least some capital ship programs – proved possible for a while.
Nevertheless, the interwar system did ultimately collapse under the pressure of broader competitive dynamics as the world girded itself for another conflict against the revisionist aggression from Nazi Germany and Imperial Japan. This “problem of revisionism” was illustrated in the interwar period by the “General Act for Pacific Arbitration of International Disputes” drafted by the League of Nations in 1928, which set out procedures for formally determining the existing legal rights of disputing parties, but which proved helpless in the face of revisionist powers’ efforts to change the geopolitical status quo.
A degree of military-geopolitical revisionism by the rising naval power of the United States, in its quasi-isolated continental fastness, was ultimately accommodated as Britain conceded the principle of overall naval parity. At least some of the German rearmament made all but inevitable by Versailles was also handled, for a while. Yet Imperial Japanese and Nazi German hegemonism and aggression in their own regions proved too much for the international system to bear – though arms limitation negotiations in some respects dragged on well beyond their actual utility as a result of fears among leaders of the major democracies of being depicted as “against” arms limitation.
To be sure, it is notable that poison gas was not used as a weapon of war in 1939-45, despite the absolutist nature of that conflict. All the same, the Second World War still turned out to be even bloodier even than the first – while other previously-agreed restrictions clearly did not hold, such as the Washington Conference’s agreement that submarines should never be used as commerce destroyers.
The interwar period thus stands as an illustration of how, while arms control can do some real good in helping mitigate the dangers presented by militarily-competitive geopolitics, such palliative care is likely to be at best temporary in the face of deteriorating strategic circumstances driven by broader dynamics of shifting power, revisionist aspirations, and technological change in the methods of warfare.
C. Postwar Efforts
The big new addition to the “How might we regulate novel and disruptive military technology?” file after
World War Two, of course, was the advent of nuclear weaponry – followed, before too long, not just by long-range strategic bombers but by ballistic missile delivery systems that eventually bought the entire globe within reach of such destructive technology in minutes. Efforts to limit nuclear weapons technology date back to a U.S. proposal for international control, in the form of the Baruch Plan
presented to the United Nations in 1946, at a time when no one besides the United States possessed the atomic bomb. This plan was rejected by the Soviet Union, however, and the competitive dynamics of Cold War rivalry between the postwar superpower blocks thereafter precluded negotiated arms limits between Washington and Moscow for many years.
It was not until the early 1970s that it was possible to reach bilateral superpower agreement upon at least some such controls, with the Strategic Arms Limitation Treaty
and the Anti-Ballistic Missile Treaty, and even then, these instruments were principally limitations on the permissible scale
of armament rather than upon its type per se. The Biological Weapons Convention of 1972
did prohibit that entire category of weapon, but this did not prevent the Soviet Union from unlawfully retaining its biological weapons program. (Russia continues to have an illegal biological weapons program to this day, for work on which the U.S. Government sanctioned three Russian research institutes in August 2020.) The Nuclear Nonproliferation Treaty of 1968
took a significant step in prohibiting the expansion of nuclear weapons capabilities to additional states, moreover, but it did not address the large – and in most cases, then still growing – nuclear arsenals by then already held by the major powers.
Significant additional steps – and in particular the actual dismantlement
of existing arms capabilities – had to await the easing of Cold War tensions in the late 1980s, and the ultimate collapse of the USSR. Specifically, the Intermediate-range Nuclear Forces (INF) Treaty
of 1987 resulted in the elimination of both superpowers’ missiles of that class, while the Strategic Arms Reduction Treaty of 1991
and the Presidential Nuclear Initiatives (PNIs) saw the former Cold War rivals begin to cut back the enormous nuclear arsenals they had accumulated during their years of geostrategic competition. (The post-Cold War era also saw the negotiation of the Chemical Weapons Convention
prohibiting chemical arms, as well as the indefinite extension of the Nuclear Nonproliferation Treaty and an effort to ban the testing of nuclear weapons and negotiate a treaty prohibiting the further production of fissile materials for nuclear explosive purposes.)
II. Modern Dilemmas
Today, much of that post-Cold War momentum seems to have run out. Two further steps were, of course, taken with the Moscow Treaty of 2002
and the New START agreement of 2010. The former, however, merely codified additional reductions that both Washington and Moscow had already independently decided to do for their own programmatic and operational reasons. New START, moreover – though recently extended for five years – originally won U.S. Senate approval only on the understanding that no further
American nuclear cuts would be permitted if Russia did not cut its large arsenal of Non-Strategic Nuclear Weapons (NSNW). In the ensuing years, however, Russia has only increased
its stockpile of such devices, including through the deployment of battalions of intermediate-range missiles the Kremlin built in violation of the INF Treaty, collapsing that instrument. (Making matters worse, the Russian NSNW arsenal achieved its present numerical advantage because Moscow refused to dismantle these weapons pursuant to the PNI promises it made to the United States
in the early 1990s, while Washington followed through with its own pledges and hugely reduced American NSNW holdings.)
Today, the U.S. and Russian arsenals have now eliminated the surplus weapons that it was felt had been made unnecessary by the easing of tensions at end of the Cold War. But tensions are now worsening, rather than improving. Today, the global environment is stressed by Russian and Chinese revisionism, military build-ups – including, in both cases, expansion of their nuclear arsenals – the retention of unlawful chemical and biological weapons programs, and provocative behavior that is driving global politics once again along an increasingly fierce competitive path. Arms reductions in this new context would clearly be much more difficult than before even if Moscow hadn’t
established a woeful track record of violating arms control treaties
and systematically undermining European arms control instruments, and even if Beijing didn’t
sneer at even talking about arms control in the first place.
As things now stand, Russia has been not only increasing its already considerable numerical advantage in NSNW but also building strategic systems not covered by any existing agreement. Meanwhile, China continues rapidly to expand the size of its own nuclear arsenal, even while working rapidly to build a new plutonium-reprocessing capability
that would allow it – if such material were used for weapons purposes – to expand this arsenal much further and faster still. Both countries, moreover, are investing heavily in new capabilities on the cutting edge of military technology, even while stepping up conventional military threats against their neighbors.
Ultimately, it has become clear that neither of these powers actually wish
a continuation of the relatively benign strategic environment created by the end of the Cold War. To the contrary, they feel the peaceable post-Cold War order slighted and disadvantaged them, and they have both set it as a strategic objective for many years now to change
that order to their advantage. Arms control and disarmament thus today face a new “problem of revisionism” that – rather depressingly – scholars of the interwar period may not find entirely unfamiliar.
Nevertheless, precisely because
the strategic environment is today increasingly fraught and competitive – and with the roster of top-tier players having now obviously expanded to include an ever wealthier, more technologically sophisticated, and better armed new superpower in the People’s Republic of China – the instinct to try to regulate or prohibit forms of warfare or weapons technology remains strong in both international and domestic politics, and such questions remain important subjects of international diplomacy and debate. This is the case, moreover, not merely with arms that diplomats have tried to limit in the past (e.g., nuclear weaponry), but also with new and emerging potentially weapons-relevant technologies, including in the novel battlespace domains of cyberspace and outer space.
III. The Need for Caution
To be sure, some such proposals for restrictions on future warfighting capabilities are simply disingenuous hokum. These are, in effect, no more than a form of what is often now called “lawfare,” in which legal argumentation (or at least pseudo-legal argumentation) is “weaponized” in order to disadvantage an adversary, particularly in circumstances in which that adversary is more legally scrupulous and more likely actually to obey legal strictures than the party engaging in such “lawfare.”
Into this disreputable category, for instance, fall various Russian and Chinese proposals in recent years for “arms control” in outer space
or cyberspace. These efforts follow in the Soviet Union’s longstanding tradition of using arms control and disarmament rhetoric as a way to mask its own malevolent intentions, to create political divisions within democratic societies and undermine Western support for deterrence and collective security, and to tempt Western leaders into agreeing to rules that asymmetrically disadvantage them – either because the terms of such initiatives had been carefully skewed for this purpose, or perhaps simply because the Kremlin would cheat.
As a particularly humorous recent example, Russia has been promoting a “No First Placement” initiative at the United Nations against placing weapons in outer space, even as it has itself
been testing anti-satellite weaponry in orbit. (As revealed publicly in the U.S. Intelligence Community’s Annual Threat Assessment
report for 2021, China also
“continues to … field” both “ground- and space-based antisatellite (ASAT) weapons.”) Such hypocrisy from Moscow and Beijing is quite breathtaking, and these diplomatic efforts are obviously little more than contemptible traps for the unwary and credulous. In my previous role as an arms control diplomat, it was an important part of my job to push back against such gambits.
Yet there is also no shortage of sincere and well-intentioned efforts to regulate or prohibit weapons applications of technologies that are felt likely to be of great – and perhaps quite disruptive – significance in future warfighting. Not all of these efforts are particularly well thought through or likely actually to do much good, and some may not even really be intelligible at all except merely as shallow virtue signaling.
Nevertheless, it has long been U.S. policy to be open to, and indeed to seek, arms control where it serves U.S. and allied security interests, and where it is likely to be verifiable and where our counterparties are likely to comply with their obligations. It is thus important to evaluate such ideas on their merits, so that we don’t miss a genuine chance to reduce risks, and to mitigate at least some of the potential impact and horror of future warfare. The trick, of course, is figuring out which proposals represent “wheat,” and which ones are merely “chaff.”
IV. Challenges of Artificial Intelligence and Autonomy
One area in which there has been much interest in exploring some form of arms control limitation is in the application of Artificial Intelligence (AI) in warfighting, particularly in connection with potentially autonomous weapons. In this field, there is much serious interest by some important Western governments, but also all too much unhelpful hysteria by some civil society activists – often associated with the so-called “Campaign to Stop Killer Robots.”
One of the first State Department policy papers I wrote when performing the duties of the Under Secretary of State for Arms Control and International Security was on the topic of Lethal Autonomous Weapons Systems (LAWS), and the challenges of carefully thinking through questions related to their potential regulation. I would refer you to that paper – the second in our “ACIS Papers” series, published in February 2020
– for the details of my take on those topics. For present purposes, however, it is enough to note that LAWS and machine autonomy questions remain a “hot” topic, with uncertain implications for the future of Defense Department weapons development and warfighting techniques depending upon whether, or the degree to which, future U.S. leaders look sympathetically upon the idea of such restrictions.
My own instinct is that this arena is not yet well enough developed or understood to permit much by way of effective regulation. This would likely be true, in fact, if it were possible adequately to define the problem set in the first place – or if it were more possible to trust that regimes such as those of Vladimir Putin in Russia and Xi Jinping in China could be relied upon to follow such rules in the absence of any viable concept by which rules restricting weapons autonomy could possibly be verified and compliance enforced. I’m open to ideas, but count me skeptical.
But lest one entirely give up hope for legal and ethical military AI, thereafter lying awake at night worrying about some future “Skynet”-type system as evoked in the Terminator
films, I’d suggest that there is still a great deal of good work being done to prevent such calamities. It’s worth remembering, for instance, how much careful legal oversight is already applied – at least in the United States, at any rate, though I have seen no evidence that either Russia or China care much about such things – to such questions of military-technological novelty.
International interlocutors who fear autonomous weapons being left in some kind of lawless, “Wild West” status, for instance, need to be reminded that U.S. practice in this area is already exemplary. As I recounted in my February 2020 paper, for instance, it is longstanding U.S. Defense Department policy – and indeed, as the Pentagon’s voluminous Law of War Manual
makes explicit, an actual requirement of international law – to review the legality of any new weapons before permitting their deployment.
In U.S. practice, therefore, all new weapons systems
“already undergo careful pre-deployment weapons reviews in order to ensure that they can be used consistent with obligations under international humanitarian law. U.S. reviews consider a broad range of factors, including how the weapon will be deployed, its intended operating environment, the concept of operations underlying the system, the tactics, techniques, and procedures (TTPs) developed for its employment, and the rules of engagement (ROEs) envisioned for it. Should any such potential weapon system fail review, it would not be deployed, or it would be redesigned until it passed muster.”
The Law of War Manual, moreover, also makes clear that LOAC rules already apply to potential future autonomous systems, and that allocating some merely factual
functions to an autonomous system – such as identifying a military target such as a tank or combat aircraft from a complex feed of real-time, multi-sensor inputs – in no way relieves a country of the obligation to ensure that any resulting attack decisions remain consistent with law-of-war principles such as military necessity, proportionality, distinction, and humanity. As the Manual
explains,
“… [I]t is not the case that the law of war requires that a weapon determine whether its target is a military objective. Similarly, the law of war does not require that a weapon make other legal determinations such as whether an attack may be expected to result in incidental harm that is excessive in relation to the concrete and direct military advantage expected to be gained. The law of war does not require weapons to make legal determinations, even if the weapon (e.g., through computers, software, and sensors) may be characterized as capable of making factual determinations, such as whether to fire the weapon or to select and engage a target. Rudimentary autonomous weapons, such as mines, have been employed for many years, and there has never been a requirement that such weapons themselves determine that legal requirements are met.
“Rather, it is persons who must comply with the law of war. For example, persons may not use inherently indiscriminate weapons. In addition, in the situation in which a person is using a weapon that selects and engages targets autonomously, that person must refrain from using that weapon where it is expected to result in incidental harm that is excessive in relation to the concrete and direct military advantage expected to be gained. In addition, the obligation on the person using the weapon to take feasible precautions in order to reduce the risk of civilian casualties may be more significant when the person uses weapon systems with more sophisticated autonomous functions. For example, such feasible precautions a person is obligated to take may include monitoring the operation of the weapon system or programming or building mechanisms for the weapon to deactivate automatically after a certain period of time.”
Machine autonomy, in other words, is no excuse for breaches of the laws of war, and strict legal accountability will be expected – under well-established legal principles – for any problems. LAWS, as it were, are already
regulated by the laws of war, and misuse in ways prohibited by International Humanitarian Law can already constitute a war crime.
On top of this this legal foundation, the United States is also leading the way in developing ethics
for military AI. In particular, after 15 months of consultations with leading AI experts in industry, academia, and government, our Department of Defense adopted in last year a set of five ethical principles for the use of Artificial Intelligence, to apply in combat and non-combat applications alike. As announced by Defense Secretary Esper in February 2020, these principles are as follows:
- “Responsible. DoD personnel will exercise appropriate levels of judgment and care, while remaining responsible for the development, deployment, and use of AI capabilities.
- “Equitable. The Department will take deliberate steps to minimize unintended bias in AI capabilities.
- “Traceable. The Department’s AI capabilities will be developed and deployed such that relevant personnel possess an appropriate understanding of the technology, development processes, and operational methods applicable to AI capabilities, including with transparent and auditable methodologies, data sources, and design procedure and documentation.
- “Reliable. The Department’s AI capabilities will have explicit, well-defined uses, and the safety, security, and effectiveness of such capabilities will be subject to testing and assurance within those defined uses across their entire life-cycles.
- “Governable. The Department will design and engineer AI capabilities to fulfill their intended functions while possessing the ability to detect and avoid unintended consequences, and the ability to disengage or deactivate deployed systems that demonstrate unintended behavior.”
V. Making Progress
Frankly, the best way at this point to help ensure that future AI use in combat is both legal and ethical is probably not
to spin our collective wheels trying to craft some kind of elaborate (and likely unworkable) new treaty regime. If you ask me, it would be better to start simply by trying to hold all
countries – including especially Russia and China, who appear to believe that AI-enabled warfighting is one of the keys to an emerging “Revolution in Military Affairs” but who aren’t exactly known for their legal scrupulousness – to the standards to which the United States already holds itself.
(I could even imagine the negotiation of some kind of international instrument – perhaps a treaty, or a legally-binding resolution of the United Nations Security Council under Chapter VII of the U.N. Charter. Such an instrument might both reaffirm
that the laws of war still apply to novel technologies in conflict, that each nation will be legally accountable for how AI is used by its armed forces, and that all states must conduct scrupulous pre-deployment legal reviews, and also require
all states to make public detailed information about the procedures they employ in complying with these requirements.)
Of course, none of that would prevent a revisionist scofflaw dictatorship such as the Putin or Xi regime from simply going through the legal motions and deploying whatever creepy autonomous monstrosity it wanted, irrespective of whether it was truly legally compliant. I also find it hard to imagine how one could have a meaningful verification regime in this arena.
That said, since we in this great Republic are going to hold ourselves to exemplary legal standards either way, we might as well “lean in” to this issue, using our scrupulousness as the legal and moral strength that it is. To my eye, therefore, the legions of civil-society activists keen to prevent the advent of “Killer Robots” should be pressed to direct their energies where they are most needed: at demanding that the world’s dictatorships demonstrably come up to the high legal standards already
being set by those of us in the advanced democracies.
While we are at this diplomatic task abroad, moreover, I would also suggest that we double down at home on ensuring that we ourselves approach military AI issues – as a matter of national policy – as wise and insightful stewards of advanced technology as this field develops. There is presently a great deal of work being done on AI issues in U.S. private industry, by some of the smartest and best-funded people on the planet. Nevertheless, it can hardly be overemphasized that warfighting-related AI applications present qualitatively different security, safety, and reliability challenges than the ones at which most industry research is apparently directing itself in the purely civilian and commercial arena.
It is critical, for instance, to ensure that any
AI applications are free of inadvertent problems caused by limitations in the data sets used to “train” its algorithms, and to ensure that such computing is secure against intentional manipulation. The appropriate standards for security and reliability, however, will be far more stringent in applications related to warfighter decision-making than in most civilian use cases. And this suggests an important national research agenda.
If we are to have what I have heard called “justified confidence” in the security, reliability, and safety of AI algorithms used in connection with future warfighting, in other words, we will have much work to do on the cutting edge of AI science and its applications in the years ahead. We will need, moreover, to do this on the front end
of the AI revolution. (As one pathbreaking AI researcher of my acquaintance pointed out to me recently, we have learned from unpleasant experiences with the Internet that retroactively
engineering proper safety and security into a system that was designed solely for speed and efficiency is far more difficult, painful, and risky than simply having designed in security from the outset.) The time to start getting “justified confidence” right with national security-related AI is therefore “now,” not “later.”
VI. Conclusion
So while I believe the right answer right now is not
to try to approach the protean and embryonic arena of machine autonomy through the traditional legal-prohibitory lens of traditional arms control, there still is important and promising work that remains to be done in helping address the legal and ethical challenges presented by AI as a novel and potentially disruptive military technology. Nevertheless, not everyone is likely to agree with me on this, and there remains much interest in some quarters for just such a “define a problem and ban it” approach.
Whichever approach one takes, however, I hope it can be agreed that we need more thoughtful engagement about these matters among relevant stakeholders. Activists, for example, tend not to appreciate the technical and operational complexities that are often actually involved, approaching such questions from a perspective that is stronger by way of passion and moralistic fervor than technical acumen. Yet the technical community in such emerging areas is unaccustomed to thinking about technology control and arms limitation at all, while very few people indeed seem to have much appreciation for whatever lessons can be drawn from the many efforts that have been made in the past to regulate disruptive and emergent military technologies.
I suppose what I’m saying is that we can all do better in thinking about these challenges. There needs to be more dialogue on such topics, and what we need most right now is less some kind of international convention than simply to spend more time building our understanding of these complicated issues – and of all those details where the Devil surely lurks.
I fully agree there can indeed be danger in failing to regulate what should be regulated. Yet there can also be danger in regulating badly, in falling into traps that someone else has set for us, or simply in trying to impose controls upon what may be fundamentally uncontrollable – or in circumstances in which such controls are, for some other reason, unlikely to work.
There is no foolproof way to prevent any of these problems, of course. But neither is there any excuse for failing to approach them with as much care and thoughtfulness as possible. As you work on whatever technology issues seem most relevant from your perspective in the Air Force Futures Office, therefore, I would urge you to work as closely – or at least as closely as classification will permit – with other participants in the policy community to ensure as much sharing of information as possible about these kinds of questions.
The better those of us in policy, legal, and political circles actually understand the technologies involved – and their real-world use cases – the less susceptible we will be to trying to impose poor answers. Similarly, the more that technology innovators like you understand about the broader arms control policy world, and the history and trends therein, the better you will be at building sustainable solutions. And with all this, the more effectively we will hopefully be able to work together in answering the mail, not just with regard to future warfighting, but also with regard to law and ethics.
Thanks! I look forward to our discussions.
-- Christopher Ford
Below appears the text upon which Dr. Ford based his remarks to the Center for Strategic and International Studies (CSIS) Project on Nuclear Issues (PONI) “PONI Scholars” group on March 28, 2024.
Dr. Ford's paper "Nuclear Posture and Nuclear Posturing: A Conceptual Framework for Analyzing China's Nuclear Weapons Policy" was published in February 2024 by the National Institute for Public Policy . You can read the paper on NIPP's website here , or use the button below to download a PDF.
Below is the text of Dr. Ford's comments at an event the American Enterprise Institute on February 13, 2024, on U.S. outbound investment screening.
Below are the remarks Dr. Ford delivered at Columbia University’s School of International and Public Affairs on February 8, 2024.
For a roundtable on December 13, 2023, sponsored by the Society for Risk Analysis and the Stimson Center , Dr. Ford participated in a discussion with Stimson's Debra Decker about nuclear risk reduction and the challenges of leadership in a complex national security environment. You can find materials on the roundtable here , and a video of Dr. Ford's discussion with Ms. Decker here .
Below is the prepared text upon which Dr. Ford drew in making brief remarks at the Carnegie Endowment for International Peace’s “Targeting Workshop” on January 12, 2024.
With 2023 now in our collective rear-view mirror, I thought I’d offer you a handy compilation of my public work product from the last year. The list is heavy on strategic competition with China, of course, but doesn’t omit other topics ( e.g., morality and nuclear weapons policy, nuclear nonproliferation, and North Korea). Keep checking New Paradigms Forum for new material as we move into 2024!
Below are the remarks delivered by Dr. Ford at the “Strategic C ompetition Educators Conference” held on December 7, 2023, at the U.S. Foreign Service Institut e in Arlington, Virginia.
Below are the remarks Dr. Ford delivered at a conference sponsored by the Center for Global Security Research (CGSR) at the Lawrence Livermore National Laboratory (LLNL), on December 5, 2023.
Below are the remarks Dr. Ford delivered at Bacon House in Washington, D.C., on October 6, 2023, to DACOR ’s annal conference. This text has been supplemented with amplifying references to the original (longer) text Dr. Ford prepared for the event.