Leaks are guaranteed---empirics prove.

Both parties might engage in secret reassurance in the face of domestic political opposition. This scenario resembles “mutual assured destruction” (MAD) in which the risk of punishment deters both sides from defecting. From the perspective of the initiator, this scenario is preferable to one in which he alone faces domestic punishment. Nevertheless, according to my proposed mechanism, the adversary should not see the secret channel as credibly revealing the initiator’s cooperative intentions, as the adversary is likely to be self-deterred from exposing. Yet, even in these MAD-like situations, the secret channel may be credible because a third party, such as journalists, mediators, or disgruntled insiders, can reveal the secret negotiations. The risk of such a disclosure, in other words, “must come from somewhere outside the threatener’s control . . . [I]t is an ingredient in the situation” that neither side can “entirely control.”40 Third-party actors may intentionally reveal information about secret talks and concessions because of ideological or political convictions, or they may unintentionally leak information without realizing the potentially adverse consequences. The level of risk may be higher for democratic initiators because they have less control over the media and their countries likely contain more players with access to the secret information.

The analogous situation in deterrence theory seems to be what Schelling calls the “threat that leaves something to chance” that creates what Robert Powell describes as “autonomous risk.” This element of risk exists beyond the control of the parties involved, suggesting that the magnitude of the chance taken affects the credibility of the action. As Powell explains, “[t]he expected cost of taking a step-that-leaves-something-to-chance varies with the level of the autonomous risk that taking the step creates. If, therefore, the risk is sufficiently small, the expected cost of taking the step may be less than the expected cost of submitting.”41 When the risk of undesirable disclosure is high, the credibility of the initiator’s signal for cooperation is stronger. Accordingly, leaders who undertake secret reassurance do so when their intentions to cooperate are sincere. Indeed, the risk of revelation from such exogenous actors is absolutely necessary in these MAD-like situations in order to make the logic of secret reassurance work from the perspective of the adversary. Incurring such risks reveals leader types because an insincere leader will not risk the costs of secret reassurance, given that other actors beyond his control can impose significant political costs on him at any point through revelation.42

The danger of autonomous leak is real. As a recent example, in May 2011, the Arabic language news network Al Jazeera released 1,600 leaked documents detailing nearly a decade’s worth of secret negotiations between Israel and the PLO. These papers, according to Al Jazeera, “show the PLO offering concessions that would have been overwhelmingly rejected by the Palestinian public, and receiving nothing in return from an Israeli government that budged little on core issues.”43 That episode significantly embarrassed the Palestinian President, Mahmoud Abbas, and the PLO more generally, leading the chief Palestinian negotiator, Saeb Erekat, to resign. To be sure, Hamas officials hailed the release of the papers, calling the negotiations treason, “a betrayal of precious and long-held national goals.”44 Similarly, the recent WikiLeaks revelations of numerous State Department cables pointing to the existence of secret agreements led to increased fears that foreign leaders may no longer enter secret negotiations with the United States in the future. Nevertheless, from a signaling perspective, if an adversary and the initiator both prefer private negotiations to preserving the status quo (and both share significant opposition to negotiations), the autonomous risk of a leak may not only reveal cooperative intentions, but may also incentivize both sides to reach swift settlements in order to avoid revelation by exogenous actors.