Bilateral Consequences of Compromised Intelligence Operations, 1985-2020

Executive Summary

It is a staple of screenwriters and novelists, subject of stock news footage, and a spectre haunting the dreams of ambassadors: a spy, unmasked, probably handcuffed, bundled between burly thugs; riots outside embassies with flags and effigies alight; newspaper headlines blaring outrage; and the chilling words persona non grata. Another intelligence operation compromised, and a carefully crafted diplomatic relationship ruined. To what extent, however, are these fears borne out? 

This report systematically analyses the impact of 174 historical cases of compromised intelligence operations from 1985 to 2020 and their real-world impact on diplomatic relations. It reveals that the bilateral consequences for states caught in these exposures have actually been much less serious than might have been expected. 

Catastrophic consequences, where the operation’s compromise drastically changed a bilateral diplomatic relationship, were not present. Critical consequences, involving major damage and pervasive deterioration of a relationship, appeared in only 10 cases. Significant consequences, with moderate but recoverable impacts, only occurred in 16 cases. All other cases demonstrated minor or negligible consequences for the exposed state.

What is also clear is that certain identifiable factors at play in these cases have more significant explanatory utility as to outcomes than certain other factors.

Factors Influencing Diplomatic Outcomes:

1. The nature of the compromised operation (namely, its egregiousness) correlated strongly with consequence. Operations characterised by violence (whether intentional or unintentional) and material damage (including cyber effects), involving covert influence/interference, or personally impacting national leaders, correlated particularly strongly with negative bilateral outcomes.
2. Less closely correlating but still noteworthy was where the target of the foreign intelligence operation was historically and/or culturally predisposed to be relatively sensitive - to the protagonist or to intelligence operations themselves.
3. Pre-existing relations were also influential. Spying on allies was risky, although not as risky as might be supposed. At the other end of the spectrum, surprisingly, transgressing on adversaries proved less damaging than doing so to neutrals.
4. Protagonists’ responses played an important and counter-intuitive role in the consequence. A strategy of unrepentance proved safer than vociferous denial.
5. Timing was more ambiguous. Compromise at politically or diplomatically inopportune moments could prove problematic. But on some occasions that inopportunity also generated mutual incentives to downplay incidents. On other occasions it was difficult to separate out the influence of timing over consequence from the circumstances of timing precipitating the activity itself (eg heightened tensions might exacerbate fall-out, those tensions might also be the reason for the intelligence operation).

Less Influential Factors:

1. While the degree to which the compromise was accorded publicity correlated tightly with consequences, publicity was almost always a function of consequence, rather than vice versa.
2. The power dynamic between protagonist and target had significantly less explanatory power. 

These findings can aid intelligence agencies, and the policymakers and envoys they serve, in sharpening approaches to risk assessment and management. For example, the findings on timing reinforce the value of a continuous assessment approach to long-term operations, rather than ‘setting and forgetting’ after initial risk assessment. The findings also have value for intelligence studies as a discipline, reinforcing the central importance of understanding intelligence as a dynamic international competition on the part of sovereign states for information and action advantages. 

Note on Timeline

The original research project was conducted during the 2019-2020 academic year at Harvard University, analyzing cases up to and including April 2020. Completion of the project was delayed by several factors, including the COVID-19 pandemic. I have since tested the original conclusions against six new cases of compromised intelligence operations after 2020, applying the same assessment framework, and compared the conclusions reached to the hypothesis first developed in 2020. The results confirm the initial findings regarding factors with greater or lesser explanatory utility in understanding the bilateral consequences of compromised intelligence operations.

Risks of Compromised Intelligence Operations

When intelligence operations are compromised and appear all over the news, it is easy to assume they will have profound negative impacts on relations between the countries involved. For instance, the twelve months following June 2013 were certainly an annus horribilis for the US and its closest intelligence allies, battered on an almost daily basis by the disclosures made by former National Security Agency (NSA) contractor Edward Snowden. The Snowden disclosures were an ambassador’s nightmare come to life. Relationships soured with diplomatic partners from Berlin to Brasilia, as revelations of intelligence operations seemed to convulse the globe. 

Intelligence professionals – and the policy makers and envoys they serve – are, indeed, preoccupied by the possible bilateral relations consequences of compromised operations, which may range from sanctions, expulsions, or other adverse diplomatic actions intended to send a message to the perpetrating state. These consequences, whether related to one’s adversary, interlocutor or ally, are at the heart of intelligence planning and risk assessment in capitals across the world. 

Illustrating the concern, the following risk assessment matrix (or something very similar) would be a fixture in agency boardrooms and ministerial suites, referenced as operational risks are thrashed out in the pursuit of national interests and priorities.

Figure 1: Typical Intelligence Operations Risk Framework

Yet, for all the preoccupation of practitioners, and the centrality of the concept of compromise to cultural and historical understandings of intelligence, this question of risk and consequence has been largely ignored by intelligence studies to date.

One needs to delve back to 1976 for Herbert Scoville’s Foreign Affairs piece ‘Is espionage necessary for our security?’ for a seminal examination of the risks to policy interests arising from intelligence compromise. Writing at the height of the post-Church Committee hysteria about US intelligence, Scoville canvassed myriad injuries done to US foreign policy by revelations of failed assassinations and over-zealous counter-intelligence. Amusingly, from our perspective post-2013, Scoville carved out from his concerns signals intelligence (SIGINT) operations, arguing that their risk profile contrasted favourably with the “political liabilities” of their human intelligence (HUMINT) counterparts.¹ Yet even in this full-throated critique, Scoville was unable to specify what the actual international political risks flowing from intelligence operations might be, or even link to specific examples (beyond reference to covert action in Chile).

It is unsurprising that more general accounts of intelligence have therefore also failed to grapple with the question of consequence. Walter Laqueur’s esteemed and popular The Uses and Limits of Intelligence,² for example, referred only to the budgetary costs of intelligence operations and agencies. 

I argue that this gap reflects the outsider’s non-recognition of the fundamentally competitive nature of intelligence. Typically, this phenomenon is dismissed as the juvenility of ‘spy v spy,’ when it is actually fundamental to understanding what intelligence is and is not. Intelligence is a competitive activity between sovereign states in their clandestine pursuits of information (and covert pursuit of action) to provide decision advantage over each other. At the core of that dynamic is the constant juggling of reward and risk.

Understanding this critically important part of the intelligence world is not trivial. Members of the public interested in international affairs should be aware of the informed attention paid by officials, acting in their names, to the risks inherent in the conduct of intelligence.

Understanding the Problem

How can one approach a topic for which the academic literature is so limited? And most importantly, how should one properly assess the actual relationship between intelligence compromise and resulting bilateral relations consequences? Are the intuitions used commonly by practitioners and policymakers valid: beware enraging the powerful; spying on allies is worse than spying on enemies; minimise publicity; avoid antagonising responses when caught; and the ‘worse’ the act itself, the ‘worse’ the consequence?

To answer these questions I chose to canvass a broad sample of cases of intelligence compromise – reflecting differing types, consequences, places, and actors – stretching from 1985, the infamous ‘Year of the Spy’ (in which numerous espionage events unfolded publicly in the US, the Soviet Union and elsewhere), until April 2020. 

In doing so I was conscious of two critically important caveats:

• These cases risk sample bias, because outside of classified discussions it is simply not possible to identify and analyse cases that have not been publicised. Importantly, publicity is not necessarily typical of compromised intelligence operations, which are very often dealt with behind closed doors between nations or even consciously ignored by the target of the activity (beyond actions directly aimed at perpetrators). Authoritarian states are particularly disinclined to air their dirty linen – fearing that the ‘success’ of catching a spy will not make up for the loss of face in suffering one.
• This is not, nor does this pretend to be, an exhaustive or exclusive catalogue of cases between 1985 and 2020. This is a sample, intended to inform a heuristic understanding. 

The methodology used in assembling the case file and categorising the cases themselves is detailed in the working notes at Appendix A. In short, I outlined the circumstances of each case; its protagonist and target; and categorised each using set scales for:

• Relative power between target and protagonist.
• Pre-existing relationship between the parties.
• Degree of publicity concerning the case.
• Nature (egregiousness) of the intelligence operation itself. Namely the transcending of run of the mill information collection into covert interference or violence. Or the infringement of some other norm – for instance by using the birth records of a disabled or deceased child for the purposes of assumed identity.
• Timing, in the context of other national or bilateral events/developments.
• The response of the protagonist to the compromise of their apparent operation.
• Particular pre-dispositions or sensitivities of the target with respect to the protagonist or to intelligence operations themselves. For example, the sensitivity of a politician of East German background to being the subject of surveillance or of a Latin American public to perceived aggression by the United States.
• Finally, the scale of consequence – for which I constructed the following indicative framework (based on the typical risk assessment framework cited at Figure 1).³

Figure 2: Indicative Framework for Bilateral Relations Consequence 

The Case File

The case file, which can be found in Appendix B, includes 176 separate cases of the compromise of intelligence operations, featuring 28 different countries as protagonists.⁴

Figure 3: Case file overall summary 

Case Selection

The cases included in this study are those that reflect the conduct of foreign intelligence activities targeting a foreign country. Typically, these involve the clandestine collection of information otherwise denied to the protagonist. In addition, I have included related instances of domestic counter-intelligence targeting another country (for example, by effecting defections of foreign officials). Some, though not all, economic espionage cases were included, with selection hinging on there being an important state-on-state dimension.⁵

Perhaps more controversially, the case file also includes acts of violence, sabotage, and interference/influence by intelligence agencies in pursuit of policy objectives. This inclusion sets this exercise apart from many intelligence studies analyses in the Anglo-American tradition.⁶ This decision is deliberate, to avoid the trap of limiting our analysis to the familiar and ignoring much of the work of intelligence agencies globally.⁷

There were several criteria for exclusion. The focus is on state-on-state relations; no cases involve spying for non-state actors. Furthermore, Cold War era US-Soviet cases were excluded. I have not included incidents involving international organisations rather than sovereign states.⁸ Instances are also excluded when it is unclear whether the action was directly linked to intelligence agencies or to other arms of government (for example, the murder of Jamal Khasoggi), or where the revelations did not involve foreign intelligence (such as the Wikileaks publication of diplomatic communications and defence holdings). There are also minimal instances involving the unfortunate contemporary practice of some states in laying transparently questionable espionage charges against foreign nationals to exact leverage over other states. These are only included when the facts are sufficiently ambiguous as to leave the door open to genuine espionage having been involved.

The date of inclusion reflects the date of revelation, not necessarily the date of the activity. So, for example, it includes revelations in the 1990s of historical Russian espionage in Israel in the 1970s.

Finally, some cases are mirror images; both the ‘blow’ and the ‘parry’ are included. For example, the Larry Wu Tai Chin case inside the CIA was propelled by the defection of Yu Zhensan from China to the USA.

Before engaging with individual cases, it is noteworthy just how prevalent espionage and other intelligence operations evidently are in international affairs. This is, after all only a non-exhaustive sample of just those cases which have entered the public realm.

Analysing Consequences

Remarkably, the consequences of these compromised operations proved much more limited than might have been imagined. Figure 4 shows the key findings of consequences divided into categories. Not one of the cases is categorised as ‘catastrophic’ in consequence, and only 10 are categorised as ‘critical’. Adding in the 16 ‘significant’ cases (aggregation of ‘critical’ and ‘significant’ is used throughout this analysis to identify meaningful consequences) reveals that only 15% of the cases rose above ‘minor’ consequence. This suggests at the very outset that much of the fear of the damage that might be done to a country’s bilateral relations, as a consequence of their intelligence operations, is misplaced.

The findings also suggest that ‘catastrophic damage’, while so prominent on risk tables, is almost unheard of in this context. The few instances in the last 120 years which might be assessed in this fashion include the sponsorship by Serbian intelligence of those who perpetrated the assassination of Archduke Franz Ferdinand in 1914, perhaps the CIA-MI6⁹ sponsorship of the 1953 coup in Iran (if the consequences are evaluated in terms of the revolution of 1979); and perhaps the shooting down of Gary Powers’ U2 spy flight in 1960 given its impact on Soviet-US relations for the following decade¹⁰.

In analysing consequences, it must be acknowledged that consequences for relations between the involved state parties are not the totality of concerns that a sovereign actor might have when considering the risk of undertaking intelligence operations. The 2013 experience, for example, indicates that other international consequences can be relevant – including impact on specific policy objectives (such as the US Government’s internet freedom agenda), domestic political concerns or business costs, and the diminution of international trust more broadly.¹¹ This study focuses specifically on the consequences for bilateral relations.

Analysing Consequences by Intelligence Type

Intelligence operations have different risk profiles that matter for this analysis. Figure 5 categorises the cases by type of intelligence activity: the conduct of HUMINT collection; SIGINT and cyber activity; covert action and/or interference; and ‘other’ (including joint operations and types of technical operation enabled by HUMINT). Just 9% of the critical and significant cases were orthodox HUMINT (and none of the critical cases) – compared to 43% of those cases involving SIGINT/Cyber and 26% involving Covert Action/Interference. This finding is tempered by the fact that few SIGINT/Cyber cases of minor or negligible consequence are likely to receive any public attention – and for this reason the figures overstate the ‘danger’ associated with SIGINT/Cyber vis HUMINT. The type of intelligence activity itself is not likely to be definitive for consequence. For example, a third of the covert action/interference cases identified resulted only in negligible or minor consequences (albeit compared with over 98% of identified instances of HUMINT).

The overall low level of consequence likely reflects the persistent degree of tolerance for foreign intelligence operations exhibited by the USA – but also Russia, China and the UK. The ‘gentlemen’s agreements’ which may have once guided much interaction between US and Russian intelligence agencies¹², for example, can also be seen between the USA and France during the 1990s economic espionage escapades, and Russia and China (eg the Danilov and Macheksport cases). Although such agreements do not always hold – for example in the aftermath of the 2001 arrest of FBI agent Robert Hanssen, who proved to be working for Russian intelligence.

Cases of Critical and Significant Consequence

Figures 6 and 7 below lay out the 26 cases in which consequences were categorised as critical or significant, respectively. What do these cases have in common?

Figure 6: Cases of Critical Consequence

Figure 7: Cases of Significant Consequence

First, these cases involve a small coterie of countries as protagonists:

4 cases – USA, Russia, Israel

3 cases – North Korea, Australia, China

2 cases – UK, Canada

1 case – France 

All these protagonists have well established and active foreign intelligence programs (or are associated with a multinational program, in the case of Canada).

Most cases (17 of the 26) involved acts of violence (intended or unintended) and/or destruction (including in the cyber realm), or the targeting of the personal communications of foreign leaders. Four others might be termed intelligence ‘coups’ involving espionage of such a scale and effect as to alarm any target (ie Pollard, Mitrokhin, Hanssen, and OPM). Two specifically involved Israeli covert procurement of friendly foreign passports for other intelligence purposes.

Otherwise, common threads seem more ambiguous. There does seem a clear correlation with publicity – of the 26 cases all but five were deemed ‘very public’. So too with the egregiousness of the act – all but 9 were ‘more egregious’. Other factors, viewed from this perspective, vary considerably. 

The Snowden Disclosures

The cases arising from Edward Snowden’s disclosures deserve a specific focus, given their prominence (7 of the 26 critical or significant cases).¹³

As buffeting as the 2013 northern summer was, along with the winter that followed, even some contemporary observers were sceptical about the longer-term implications for US relationships – especially where outrage was somewhat less sincere. This was perhaps best expressed by former French Foreign Minister Bernard Kouchner: 

“Let’s be honest, we eavesdrop too. Everyone is listening to everyone else. But we don’t have the same means as the United States, which makes us jealous.”¹⁴

Or Robert Jervis’ observation that: 

“...the changes [to US bilateral relations] will be even less because there was little that the United States was doing that was really unusual, something that foreign leaders knew despite the domestic political incentives that led them to express outrage”.¹⁵

We see also some very interesting and farsighted approaches by target countries as diverse as Mexico and India, both of whom ultimately muted their responses, despite US surveillance of the newly elected Mexican President Enrique Peña Nieto being characteristic of the same circumstances which drove outrage in Berlin and Brasilia. 

Nonetheless, the prominence of the ‘Snowden cases’ amongst those with the most meaningful consequences does suggest that the phenomenon of 2013-14, compared to one-off compromises, is more than the sum of its parts. This may in turn reflect changing societal perceptions regarding privacy and communications (see conclusion below).¹⁶

Common explanatory factors evaluated

As noted above, the case file includes categorisations of the following factors for each case¹⁷ of intelligence operations compromise:

• Nature (egregiousness) of the intelligence operation itself.
• Particular pre-dispositions or sensitivities of the target with respect to the protagonist or to intelligence operations themselves.
• Pre-existing relationship between the parties.
• The response of the protagonist.
• Timing, in the context of other national or bilateral events/developments.
• Degree of publicity concerning the case.
• Relative power between target and protagonist.

The following observations are based on analysis of the aggregate data for each factor.

Nature of the Operation: Significant explanatory power

The particular nature of the operation – its egregiousness from the perspective of the target – correlates strongly with consequence. Typically, more egregious intelligence operations have been those which involved violence or loss (including unintentionally – as in the death caused by the French Operation SATANIC in New Zealand), or impacted national leaders or democratic institutions (as in Russian interference with US elections). They have also involved particular transgressions of societal norms (such as the use of the identities of deceased children or the profoundly disabled).

80% of cases with critical consequence involved ‘more egregious’ operations (reducing to 65% when including significant cases also) – compared to only 9% of less consequential cases.

One methodological limitation: there is a certain retrospective dimension to this categorisation, as an observer may be somewhat more likely to assess a case as ‘more egregious’ if the consequences were profound.

The Snowden disclosures included a number of well-known instances of espionage seen as ‘more egregious’ (at least in the eyes of the target states) – and of grave consequence - for having infringed the persons of leaders like President Yudhoyono, Chancellor Merkel and President Rousseff. However, it is worth remembering the counter-example of Mexican President Enrique Peña Nieto, where Mexico’s response to alleged surveillance was more muted.

Pre-Disposition of Target: Some explanatory power

Where a target of an intelligence operation is pre-disposed to view with sensitivity either the operation’s particular protagonist or perhaps intelligence activity itself, this can exacerbate the bilateral consequences of the operation being compromised. 

So, for example, there was a heightened reaction of Latin American governments – otherwise pre-disposed to view the US as a regional ‘hegemon’ and sometimes ‘bully’ – when allegedly targeted by US SIGINT operations revealed in the Snowden disclosures.

The results can also reflect unique historical and cultural perspectives of the targets themselves. This was particularly evident in the Snowden-inspired case of alleged NSA operations against Germany and Chancellor Merkel. To quote former Director of National Intelligence Jim Clapper:

“So, for [Merkel], the Stasi wasn’t a mythical bogeyman. She had grown up under its oppression, and for that reason, I believe she never [emphasis in original] trusted intelligence organizations – hers or anyone else’s.  She didn’t know and didn’t want to know what her intelligence services were doing, and the reports from Der Spiegel that said the BND was helping NSA spy on her and on German citizens recalled the real-life experiences of her childhood and young adult life all too well. Even worse, her experiences and biases were not – and are not – outliers among German politicians.”¹⁸

In the data a pre-disposition to sensitivity was particularly apparent in the critically consequential cases – 60% of such cases featured ‘more sensitive’ dispositions, compared to 21% of cases with minor or negligible consequences.

Pre-Existing Relationship: Some explanatory power

If power dynamics cannot effectively explain differing consequences, then surely the pre-existing relationship between protagonist and target should be able to do so? Surely spying on friends and allies, for instance, would have greater consequences than doing the same to a neutral?¹⁹

The cases instead suggest that relations might not be as influential as we imagine, and cannot explain all outcomes. Why, for example, was the Pollard affair so much more consequential for US-Israel relations than the later Kadish affair between the same two parties?

Nonetheless, the data does suggest that this factor has explanatory power. Of the critically consequential cases, 27% involved a previously allied or friendly relationship – rising to 33% when also including significantly consequential cases. In addition, 20% of the negligible or minor consequence cases involved friends/allies. However, as an effective control group, neutrals featured in these cases 18%, 10% and 22% respectively.

Importantly, 18.9% of cases involving allies/friends resulted in a consequence rating of critical or significant – rising to 21.4% if considering allies alone; compared to 6.4% involving neutrals. This suggests that spying on allies, while by no means certain to produce critical or significant consequences if compromised, is unsurprisingly more dangerous for bilateral relations than spying on neutrals – despite examples of muted reactions such as Operation SOCIALIST (alleged GCHQ operation in Belgium).

Adversarial relations are particularly interesting - 12.5% of the cases involving adversaries resulted in critical/significant consequences, increasing to 16.5% when also incorporating cases with ‘complicated’ pre-existing relations. The limited degree to which such relations can deteriorate further is evident in the particular cases involving India-Pakistan and Israel-Iran, for example.

Protagonist’s Response: Counter-intuitive findings

A protagonist’s initial response can also play a part in de-escalating (or alternatively escalating) the bilateral consequences of a compromise. The classic neutral response of ‘neither confirm nor deny’ featured in just 22% of critically consequential cases (27% when critical and significant) – compared to 49% of less consequential cases. Denial seems like a particularly dangerous strategy for an accused protagonist: such a response features in 68% of critically and significantly consequential cases, and only in 41% of less consequential cases.

Surprisingly, a strategy of unrepentance seems less hazardous than vociferous denial. In 8% of less consequential cases the protagonist chose to brazen out the offence.

Perhaps the most ham-fisted protagonist response was North Korea’s 2002 less-than-apologetic attempt to address decades-old rumours about the kidnapping of Japanese citizens by Pyongyang’s intelligence agencies. In another example of being dragged along by media reporting (and resulting opinion) the admission ended up compelling Tokyo into a series of robust sanctions against the Hermit Kingdom.

Underlining the broader point about the implicit acceptance of intelligence operations in world politics, there was only a single instance in the case file of a protagonist leading with an apology (rather than being ultimately compelled to offer one) and that involved allegations of US spying against Japan.

Timing: More ambiguity

Timing was notable in 52% of significant and critical cases, although it was absent in 83% of less consequential cases. There were also cases in which it was apparent that circumstances of timing might generate mutual incentives to mute consequences (as in 2007 when the German Chancellor was visiting Beijing) or circumstances simply drowned out consequences (as with Montes’ arrest in 2001 for spying for Cuba in the midst of US response to 9/11. 

Even in cases where timing seemed to exacerbate consequences it was difficult to distinguish the influence from the precipitation of the activity itself. For instance, in the worsened bilateral relations that affected the consequences of a ‘hack and release’ by Russia of a senior US official’s phone calls in 2014 amidst crisis in Ukraine but also encouraged the mounting of the Russian operation itself.

Publicity: A dependent variable

The data suggests a very strong correlation between the level of publicity and consequence. 83% of critical or significant cases featured a ‘very public’ level of publicity, compared to 10% of the cases with negligible or minor consequences. However, it turns out that publicity follows consequence more than vice versa; it is a dependent variable rather than an explanatory one. 

The principal distinction made between ‘public’ and very public’ in this instance is the persistence of media coverage and public discussion, beyond the initial revelation, as well as a broadened range of outlets for publicity, including internationally. 

There are isolated cases where publicity did help to drive reluctant target governments towards a more robust response to protagonists. For example, the coercive Chinese cultivation of a Japanese code clerk in Shanghai in 2004, which drove the clerk to suicide, was ignored by the Japanese Government until revealed in the Japanese media – at which point it was compelled to take actions that raise the bilateral consequence to minor. Overall, however, publicity appears to have been driven by consequence rather than the other way around.

Power Dynamic: Not a significant factor

It seems obvious that the relative power of the parties should make a difference to consequence. Surely a more powerful target is better placed to swat a weaker protagonist? Or is a less powerful target more likely to be outraged by the bullying of a stronger protagonist? However, the evidence suggests that neither is necessarily the case. 

Power is a much weaker explanatory factor than anticipated. While most critical cases (80%) involved a less powerful target, this is an n of only 10, and it falls to 55% when including cases of significant consequence.

Furthermore, cases involving parties of similar power resulted in significant or higher consequences 15% of the time – the same outcome as those involving a less powerful target (15.7%), and more than cases involving a more powerful target (10.3%).

Individual cases of negligible or minor consequence also suggest that weakness doesn’t preclude states emerging unscathed from intelligence operations compromised while targeting the powerful. Although it is certainly true that protagonists benefitted from targeting the (more tolerant) USA in particular (as in cases where operations by Ecuador, El Salvador, Greece and others targeted US interests.).

Impacts Over Time

The case file affords an opportunity to consider the question of whether the consequences of compromise have changed over time. For instance, did the Snowden disclosures bring about a world more tolerant of intelligence operations or more resigned to them?

In fact, as figures 15(a), (b) and (c) show, there has been a considerable stability in the consequences of intelligence operations being compromised. The period 2011-15 appears to be an aberration (reflecting the particular influence of the Snowden cases), with consequences of compromise in the last five years returning to the levels more common to the preceding two decades.

How have target countries adjusted after being targeted? The aggregate data is less useful in considering the question of whether the experience of being the subject of others’ intelligence operations sensitises a country or inures it. 

There does appear to have been an effect in the UK when analysing the government’s response to the 2006 Russian assassination of Litvinenko and the 2018 attempted Russian assassination of Skripal. So too, one could read into the heavier consequences of the Hanssen case the memories of Ames, Pitts, Nicholson, and others. The same is true of the more robust US response to Chinese espionage (especially cyber-espionage) since 2014, compared to the remarkable streak of muted responses to uncovered Chinese intelligence operations in the three decades beforehand. Although this could well also reflect grander strategic factors, it is worth also considering whether the cumulative effect of these instances may have had a broader deleterious effect on American public, commercial and elite perceptions of China and its rise.²⁰

Conclusion and Policy Recommendations

This analysis leaves us with several conclusions that have implications for intelligence professionals, scholars, and policymakers.

The critical finding of this study is that contrary to impressions, catastrophic bilateral consequences from intelligence operations are almost unheard of, and critical consequences are rare. In particular, the compromise of standard HUMINT operations presents little risk of serious bilateral consequences. Thus, governments should lean into the conduct of intelligence operations in world politics.

This does have limits. For example, Spanish political scientist Fernando Ntutumu Sanchis makes the important point that historical acceptance of espionage as a normal (if unacknowledged) part of international affairs is being disrupted by a perceived extension into individual lives, given the 21^st century ubiquity of digital communications within whole societies as distinct from just between states.²¹ Furthermore, this form of ‘mass su’veillance' might have the strategic effect of shifting espionage from an unconscious tool for international peace (by eliminating dangers of surprise and by reducing the security dilemma) to “what might be an epoch of international relations instability”.²²

For intelligence practitioners, Figure 16 reveals the factors that lead to significant consequences. 

This analysis suggests that bilateral relations risk assessments used in intelligence operational planning should emphasis consideration of the nature of the operation (i.e. its ‘egregiousness’ if revealed to the target) with particular sensitivity to prospects for violence and loss, effects on national leaders, and broader norms (the ‘ickiness’ factor). 

Even though timing is a somewhat ambiguous factor, prudent consideration should be given to prospective events which might accentuate consequences should the operation be compromised at that same time. In particular, this is a call to arms for a continual evaluation approach to such risk assessment. Efforts should be made to regularly review risk assessments to take into account prospective developments.

In addition, there is merit in considering the greater risk associated with activities directed against allies, while acknowledging that this is less important than the nature of the operation and its timing. Thoughtful planners should also put similar levels of consideration into any unique political or cultural dimensions of the target in the context of the particular operation.

By contrast, less emphasis should be given to power dynamics. And while no protagonist should welcome heightened publicity of their failure, such publicity is more a function of consequence than a contributor. 

This study has been limited to the direct impact of a protagonist state’s intelligence operations on its own bilateral relations with the target. In analysing the broader implications of intelligence revelations and scandals, it is important to consider the cascading consequences on the protagonist state’s international reputation. Farrell and Finnemore have argued for the debilitating effect of the Snowden disclosures on the US’s ability to deploy a foreign policy based on “hypocrisy”.²³ Easley has similarly argued that compromises erode a broader international trust, affecting sovereign states indirectly (especially when their interests are invested in maintenance of such trust).²⁴ Specific operations may have further consequences in domains such as public health, as illustrated in the case of intelligence operations that have used international vaccination projects as operational cover.²⁵

This paper leaves open significant possibilities for further analysis. The findings herein, and conclusions drawn, can be tested against those other instances which have not been included but otherwise fit the definitions used. For example, what was the effect of CIA counter-terrorism renditions on certain bilateral relationships? Or the conduct of drone strikes by intelligence agencies? 

Furthermore, there have been many more cases (and further developments in cases herein) since April 2020. Please see the postscript for a brief assessment of these cases, conducted after the formal conclusion of this Belfer Center-based research.) And from an historical perspective, might the study of cases from the early Cold War, where international relations where in relative flux, also prove useful in evaluating the findings?²⁶ There would also be utility in agencies testing the findings against publicly unrevealed, classified instances of compromised operations (whether they were poacher or gamekeeper in those instances). Another area for further study is the question of whether expectations matter: if a protagonist is regularly exposed does a version of the ‘bigotry of low expectations’ kick in on the part of those they transgress?

Finally, the findings of the paper reinforce that the fundamentally dynamic nature of intelligence should not be ignored, including in academic study. The study of intelligence is not about perennial debates on process versus product or yet another dissertation on ‘politicisation’; it’s a covert contest of capabilities and wills (and risk calculations and appetites) between states carried out for decision (and action) advantage and as an expression of contest in itself.²⁷ This study reflects how states interact when the lights are off, with consequences that are not always what one would expect.

---

Statements and views expressed in this commentary are solely those of the authors and do not imply endorsement by Harvard University, the Harvard Kennedy School, or the Belfer Center for Science and International Affairs.