Michael KreponThe Future of Arms Control

The control of weapons of mass destruction has been about numbers, treaties and norms. The first two get most of the attention, even though the third is the grand prize. Numbers and treaties are essential tools to build an edifice of normative behavior and customary practice. Numbers and treaties are the means. Norms are the ends.

Strategic arms control is a process, and is therefore numbers-oriented. Treaties relating to nuclear testing, chemical and biological weapons codify norms. As treaties become harder to enter into force, or as they enter into force without key states, customary practice becomes all important. Treaties and numbers still matter, in large measure because they reinforce norms. The number of treaty signatories is consequential even when some of them do not deposit instruments of ratification, preventing entry into force. Through customary practice, norms can become stronger over time even when treaties are stalled.

Sometimes the attention paid to numbers is misplaced, as Thomas Schelling argued and as discussed in last week’s post. Even so, numbers still have considerable import, as they provide a snapshot of the work in progress we call arms control, disarmament and nonproliferation. No construction plans are fixed and complete, despite the blueprints we have to work with.

Occasionally, construction projects require build-downs. Strengthening the global regime against nuclear proliferation is linked to responsible behavior by the two states with the worst records of nuclear excess. Reductions in nuclear force structure and inventories by the United States and Russia are necessary, but hardly sufficient to strengthen the norm of nuclear nonproliferation.

In terms of the strategic balance, it doesn’t matter whether Vladimir Putin tries to build up to New START ceilings with a liquid-fueled, MIRVed replacement to the SS-18. Nonetheless, it matters a great deal if Putin stiffs President Obama’s offer for parallel reductions below New START limits. Putin seems intent to mortgage Russia’s future in many ways, one of which is to equate national power with weapons that will not influence events in Moscow’s favor. He does, however, have the power to slow down a quarter-century-long process of parallel strategic arms reduction and corrode the bargain between nuclear haves and have-nots in the Nonproliferation Treaty – that continued abstention is predicated on the sincere pursuit of disarmament among bomb holders.

Yes, I know the counter argument: non-nuclear weapon states continue to pledge allegiance to the NPT out of self-interest, regardless of nuclear numerology. This revisionist argument is unpersuasive because it disregards the central bargain on which the NPT regime was constructed. Moreover, it can have pernicious consequences by providing cover for bomb holders to become slackers, while discounting the importance of numbers in the calculations of abstainers as to whether the NPT remains in their national security interests. Deeper cuts by Washington and Moscow are just one indicator of the continued health of the NPT regime — but an important one. Abstainers that rely on extended US nuclear deterrence do not require New START-level numbers for solace; lesser numbers will do. Even so, other corrosive developments will plague the NPT, particularly from the nuclear weapon-related programs in Iran and North Korea.

Free-riders watch on the sidelines as others engage in construction projects. Free-riders to the NPT like India and Pakistan are growing their nuclear inventories while enjoying norms that make it more costly and more obvious for new entrants into the bomb-making business. The subcontinent’s free-riders have not repaid debts to the NPT with signatures on the Comprehensive Test Ban Treaty.

China, another state with a growing nuclear inventory, has not ratified the CTBT. Beijing has benefited as much, if not more, than any state party to the NPT, while North Korea, which depends on China’s largesse, continues to grow fissile material stocks and has not ruled out additional nuclear tests. Another free rider, Israel, has refrained from ratifying the CTBT. If the United States and Russia remain stuck on New START numbers – less than three years after reaching them, with seven more years before the Treaty’s current end date – they, too, will hold passes as free riders. Free ridership corrodes norms and treaties.

The number that matters most in norm-setting is zero. This number is the clearest and most meaningful way to sets norms and customary practices among responsible states. The number zero is embedded in the CTBT, the Chemical Weapons Convention, and the Biological Weapons Convention. States that do not honor the number zero become, ipso facto, outliers. Because numbers of chemical and biological weapons that are greater than zero can be hidden, suspicions can only be conclusively affirmed by use, if they cannot be revealed by national technical means or intrusive treaty-monitoring regimes.

Norms are about behavior as well as numbers. Behavior is usually easier to monitor than numbers. The use of chemical and biological weapons becomes an indelible stain on the user. Who wants a resumé like that of Bashar al-Assad? How many states wish to join the company of North Korea in testing nuclear weapons?

Norms are also being codified in conventions against the transfer and use of certain conventional weapons, including some types of land mines and cluster munitions. These norms have important outliers, as well. Being an outlier in compacts dealing with conventional weapons entails less stigma, at least until customary, responsible practice becomes more recognizable and expected over time. This long and winding road lies ahead.

The clearest and cleanest way to codify norms and numbers is by treaties but, as noted, treaties for the foreseeable future won’t enter into force or, if they do, they will enter into force without key parties. As long as treaties remain in this limbo, success in arms control, disarmament and nonproliferation will depend increasingly on customary, responsible practices that harden over time into norms.


  1. Mark Gubrud (History)

    Numbers matter in arms control because they are perceived (rightly or wrongly) as objective measures of relative power, or of parity and equal status. Norms (especially when codified in legal instruments) that are (perceived as) clear, objective and fair are important for the same reason.

    Another norm that many of us are working to establish, even before the weapons exist that would violate it (although there are some that do already), is that the decision to initiate the use of violent force should never be made by a machine. In other words, no killer robots.

    The Department of Defense in November promulgated a highly aggressive new policy for the development, acquisition and use of autonomous weapons, including systems that would target and kill human beings without any human involvement in the decision to kill. While the policy has been widely misreported as cautious, and even a “moratorium,” in reality it marks a commitment to the full development of this technology and to its “integration into operational planning” and use in war.

    The principle that machines should not be allowed to make the decision to kill marks a bright red line that should not be crossed. If we cross this line, what further line will we not cross before armies of robots commanded by AI systems confront each other across the globe?

    • Jonah Speaks (History)

      Mark, How would you operationalize this proposed norm of “no killer robots”? I assume you do not propose to stop all research into artificial intelligence?

      I assume no nation would allow robots to determine autonomously whether to start a war, or how to wage a war once started, or allowed to continue killing after a war is officially ended. Hence, I presume you are concerned that robots might be programmed to follow a general order to kill, will autonomously select their targets subject to pre-programmed criteria, and will require no further permission to kill such targets. I presume that your proposed solution is that, after selecting a human target to kill, the robot must always request human permission before killing.

      One can imagine both hardware and software that could implement this proposed constraint. Hardware connected to the robot’s killing apparatus would require simultaneous triggering by both the AI system and the human commander. Software programming of the AI system would not permit the AI system to issue a kill command unless it too had human permission. Perhaps both hardware and software should be so designed.

      Communication with the human commander might be intermittent, and targets might not stay in the same place. For how long should human permission be allowed to activate the robot’s killing mechanism? 10 seconds? 30 minutes? 24 hours? Should prior human permission be required for each and every target? Should advance human permission be required to destroy any and all targets, both human and nonhuman? May a robot injure or hurt a human without prior permission, provided the human is not killed?

      These proposed constraints would render a robot unable to kill if it lost communication with its human commander, perhaps because of enemy jamming. What should a robot do then? Power down? Run away? Hide?

    • Mark Gubrud (History)

      Jonah Speaks,

      Sorry for the belated reply.

      No, I don’t propose to stop all research into AI.

      Why do you assume no nation would allow robots to determine autonomously whether to start a war? This has almost happened, if you consider the US and Soviet/Russian early warning systems as robots of a kind.

      In general, I don’t think any nation is eager to hand over the national command authority to a robot, but this can happen by creeping degrees and it is the logical endpoint of the road we’re now on. Already the logistical planning and scheduling for our wars – the actual marching orders – are handled by AI systems. And now that the issue of machine kill decision has been raised, already “thoughtful” people are having the epiphany that, after all, AI might turn out to be better than people at making these kinds of decisions. And maybe so. Or not.

      Your have started some good thinking about the complexities of policing the use of military robots within the constraint of no autonomous decision to use violent force. To the extent that this very clear red line imposes some difficulties and would hamper the military effectiveness of killer robots in some situations, well, exactly. Nobody ever said arms control and global security were compatible with always being able to do whatever may sometimes seem useful in war or preparing for war.

    • Jonah Speaks (History)

      The current AI for robots is more computer-like than brain-like. Each type of “intelligence” has its own strengths and weaknesses. The computer can filter out stupid or immoral actions from a strictly computational perspective that humans find difficult. The human can quickly sort through some of the nuances of complex situations that leave computers befuddled and clueless. Machine intelligence and human intelligence working together can make war (somewhat) safer by filtering out stupid or immoral actions from either perspective.

      Sensors in early warning systems provide the warning, but humans make the final decision. If launch on warning is one of the options, time pressure and human fear could cause machine error and human error to combine into an “accidental” nuclear war. The problem is the compression of time (and second-strike vulnerability), not that machines made the decision.

      Because robot AI is computer-like, it must be pre-programmed with little opportunity for learning or improvising in unexpected situations. Think Microsoft operating systems with millions of lines of code. If there is any bug in the robot code, the “blue screen of death” could take on a new meaning. Keeping humans always in the loop could prevent fatal robot errors. It also maintains continuous human responsibility for the actions of their weapons, including robot weapons.

      For a number of reasons, it makes sense to establish a norm of advance real-time human permission for all robot killings of humans.

  2. Shams uz Zaman (History)

    I would partially agree with Krepon on “The clearest and cleanest way to codify norms and numbers is by treaties… but…treaties for the foreseeable future won’t enter into force….As long as treaties remain in this limbo, success in arms control, disarmament and nonproliferation will depend increasingly on customary, responsible practices that harden over time into norms.”

    There is yet another problem of norms being broken due to over all strategic environment or security dilemmas. For instance, North Korea established a (negative) norm of breaking away from NPT, the Saudis or Egyptians, due to regional or internal compulsions, could follow suit especially if the American influence in the region continues to decline.
    Breaking a Norm is not always consequence of the state’s normative behavior or customary practices (which obviously could vary from one state to another) but rather could also emerge from irresponsible behavior of global powers. Such behavior could at times be justified under the guise of an existential/security threat or at times a quest to establish/maintain global supremacy.
    India for example was under no existential threat to develop nuclear weapons yet it provided Pakistan with a justification, Pakistan could have done well without proliferating nuclear technology but the perception of “clash between civilizations” besides monetary incentives provided a pretext for that, Iran could have continued its peaceful course but the fate of Iraq and Libya probably convinced it that nuclear weapons were crucial to deter a global power. North Korea is a complex case which could easily find internal or external pretext to develop nuclear weapons.
    Therefore, while treaties provide good framework to establish (positive) norms and practices, complex political, regional and global issues could draw a parallel (negative) trend of breaking away from these norms.
    Solving these complex issues thus would not only strengthen the arms control and non-proliferation regime, but could also ultimately lead to nuclear disarmament.

  3. Captain Ned (History)

    Zero. Total disarmament.

    Impossible fallacies propounded by Pollyannas ignorant of the real world.

    • Mark Gubrud (History)

      A large and growing number of people propose zero nuclear weapons. However, “total disarmament” would seem to mean zero weapons of any type. Which “Pollyannas ignorant” propose that?

      Misrepresenting others’ positions is not serious engagement.

  4. Bradley Laing (History)

    New Delhi/Chennai: India took a major step towards attaining its long-awaited nuclear weapons triad as a reactor on board the country’s first indigenous nuclear submarine INS Arihant was activated early yesterday, paving way for its deployment by the navy after sea trials.
    Prime Minister Manmohan Singh described it as a “giant stride” in the progress of the country’s indigenous capabilities. Defence Minister A K Antony also lauded the “tireless efforts” of the large team that had worked on the project.
    The miniature nuclear reactor on INS Arihant (slayer of enemies) went critical early yesterday and the submarine was on its own power.
    “The reactor in INS Arihant is functioning well. It will be shut down later to study the reports generated,” Atomic Energy Commission (AEC) chairman
    R K Sinha told IANS over phone.
    He said the design was unique as compared to a land-based reactor.


  5. Bradley Laing (History)

    BEIJING, Aug. 12 (Yonhap) — China has agreed to share nuclear blast data from its 10 monitoring stations with a global monitoring organization, according to a media report on Monday, in a move expected to put more pressure on North Korea’s nuclear weapons program.

    The decision was made at a meeting in Vienna between Lassina Zerbo, the executive secretary of the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO), and Zhang Yulin, a deputy minister of China’s defense ministry, Hong Kong’s South China Morning Post reported in its Sunday edition.

    China’s patience with North Korea has been wearing increasingly thin, particularly after the North’s third nuclear test in February. Beijing voted in favor of sanctions by the U.N. Security Council to punish Pyongyang for conducting the nuclear test.

    Still, it remains unknown what type of fissile material North Korea used in its February nuclear test


  6. Bradley Laing (History)

    Brazil military regime wanted the ‘bomb’ fearing an armed conflict with Argentina
    The military dictatorship that ruled Brazil from 1964 to 1985 was planning to develop an atomic bomb according to secret documents from the Armed Forces Chief of Staff to which the influential newspaper O Estado de Sao Paulo had access and released details.


  7. j_kies (History)

    If we want successful arms control; convince the ‘users’ of those arms that they do not wish to apply or face those items in conflict. Perceived self-interest will create institutional acceptance of constraints.

    Based on the history of more successful arms control protocols, e.g. the CWC and BWC and the START family; I believe protocols are successful when the military establishments consider the treaties / protocols to support their interests.

    CW and BW were shown to be indiscriminant and rather ineffective in terms of battlefield use (HE was vastly more effective) so the militaries were in favor of elimination of the logistics train of defensive measures as much as possible.

    Nuclear explosives have similar problems in use magnified by fallout and other downsides disproportional to any military ‘effect’.

    To date, the arms control community has failed to convince the using militaries that mines and cluster munitions do not provide ‘utility’ in the time of conflict or that the societal issues post-conflict trump the prompt combat utility.

    Any future arms control issue will likely experience the same acid test; “does foregoing a particular weapon enhance the military establishment” in its own view?

  8. Bradley Laing (History)

    MANILA, Philippines – The Philippine panel negotiating with the US on the increased rotational presence of American troops will deny access to US assets carrying nuclear weapons and will assert Philippine authority over the facilities of the Armed Forces of the Philippines (AFP).

    Foreign Affairs Assistant Secretary Carlos Sorreta, Philippine panel head, said mere suspicion that a US asset has nuclear weapons is enough to deny access.

    “If we suspect or believe that the ship (that) they ask for permission to enter has nuclear weapon, then we have the right to deny it,” he said.

    “We actually don’t have to see it (or) actually touch the weapon.”

    The Philippine and US panels agreed to respect the constitutional prohibition against nuclear weapons during the first round of negotiations last Wednesday, Sorreta said.