Appendix A: Forensic Analysis - How Might Intelligence on Iraq Have Gone Wrong?

Insofar as Iraq constitutes an intelligence failure, it represents a relatively unexamined variety of failure. Most events examined as intelligence failure such as the 1941 Pearl Harbor attack and the German invasion of Stalin's USSR in the same year, Rhodesia's Unilateral Declaration of Independence in 1964 or the 1968 Tet offensive in Vietnam, the 1981 Falkland Islands invasion and, of course, the September 11 2001 attacks on the United States are more properly understood as failures of warning. The failure of intelligence over Iraq was not, at any point, inadequate or erroneous warning. It was, if anything, a spurious warning based on an overly alarmist assessment which increasingly appears to have been ill supported, not merely by the facts uncovered since the invasion, but even by the intelligence and evidence available at the time.

The role of intelligence in deciding an action like the invasion of Iraq involves two very different kinds of intelligence questions. The first question is whether or not Iraq possessed non-conventional weapons, and on what scale. The second question is whether or not the available intelligence indicated that military action was the best way of dealing with the Iraqi non-conventional weapons threat. This is a significant distinction if only because of one fact that has tended to be forgotten since the Iraq invasion. That is that, as late as the autumn of 2002 Director of Central Intelligence George Tenet (whose agency was convinced Iraq had significant CBW stockpiles) was warning that Iraq did not constitute and immediate threat to United States security (Tenet, 2002). Although Tenet maintained a stance of scrupulous impartiality regarding government policy and denied any 'inconsistency' with administration rhetoric, Tenet's assessment was widely viewed as divergent from the hawkish rhetoric of President Bush and his leading advisers.

Given that there appears to have been a sizeable assessment failure, the first order of business must be to try to identify what may have gone wrong. Lord Butler will, of course, have had access to a range of secret papers such as source reports and assessments which are not, and are not likely to be, available to observers on the outside trying to make a useful analysis. However, considerable information has been made available through the Foreign Affairs and Intelligence and Security Committee reports on Iraq, the wealth of evidence made available through the Hutton inquiry, and finally selectively leaked and uncovered by the media. From this material it is possible to glean a number of hypotheses about where failure may have occurred, and why. Intelligence in this type of context may be said to be composed of four main elements: Requirements, Collection, Analysis and Estimates, and the Intelligence-Policy interface.[3] A failure of Requirements is typically when those responsible for setting national requirements and priorities for intelligence do not include a matter of significant concern which then leads to a crisis or subsequent failure of warning. For example, Argentina's relatively low priority in the UK's Annual National Intelligence Requirements Papers at the turn of the 1980s meant that inadequate resources were being expended on the region to effectively monitor Argentine capabilities and intentions concerning the Falkland Islands. A less crucial failure was the absence of Rwanda from US national intelligence priorities when America was confronted with international pressure to intervene in the internecine violence there during the mid-1990s. However, given the significance of Iraq in Anglo-American defence and foreign policy priorities after the first Gulf War, a failure of requirements seems profoundly unlikely. As far as the Intelligence-Policy interface is concerns, Lord Butler's brief explicitly omits any consideration of the political decision-making process. As a result, apart from the difficulties addressed above under 'Lost in Translation' and Recommendation 1, a diagnosis of failure at the level of policy-making is unlikely. This leaves only collection and analysis.

Collection:

One of the Butler's primary concerns is to establish the accuracy of the raw intelligence that was feeding into national assessments and decision-making prior to the invasion of Iraq. From a collection point of view, the controversy over the so-called '45 minute claim' embodies almost all of the operational problems that were confronting the intelligence community in its efforts to assess the state of Saddam Hussein's WMD capabilities. Most of the post-Cold War security priorities such as counter-terrorism, counter-proliferation, countering transnational organised crime, and peacekeeping in conflicts dominated by so-called 'ethnic entrepreneurs' and their cronies are threats generated by sub-state actors (SSAs) rather than traditional nation-states. They are typically networks of small, compartmentalised cells or associations which do not usually control significant territory and rely on stealth and concealment to conduct their activities. They are not, therefore, as amenable to technical methods like overhead imagery (IMINT) or large-scale signals intelligence (SIGINT) operations as were the concentrated, large scale threats from nation-states like Communist Russia and China or the Axis powers of the Second World War. To be sure, SIGINT remains of considerable value. As one UK official put it in the late 1990s, more terrorists and drug barons use mobile phones and the internet than ever before. However, the primary means of penetrating clandestine terrorist, criminal and proliferation networks remains the acquisition of human sources of agents inside the target organisations or affiliated bodies. From human sources one can receive eyewitness accounts of events or actions, copies or photographs of original documents, and a subjective but very often valuable insight into the interplay of personalities, agendas and orientations in a target government or organisation. A human source can be present at a meeting of terrorists at a safe house in a third country that would be inaccessible to communications intercepts or satellite photography.

Given the nature of the problem of assessing concealed Iraqi WMD capabilities, neither IMINT nor SIGINT was likely to be the main source of raw intelligence. The patchwork of overhead images, inference and supposition that made up Colin Powel's February, 2002 presentation to the United Nations Security Council is a good illustration of how little of any clarity or certainty could be gleaned from overhead imagery in the case of Iraq. Human intelligence was necessarily going to be the main source of raw intelligence about Iraqi WMD. This is in no way unique to Iraq; collection operations against other WMD-aspirant states or against terrorist groups like Al Qaeda or Islamic Jihad is necessarily likewise dependent on HUMINT. Therein lies perhaps the most persistent and challenging difficulty for effective intelligence collection in the present international crisis.

This reliance on human intelligence or HUMINT carries with it certain key risks or difficulties. Agent, that is, informants, are human beings and however high their level of access and however reliable their track-record. They are prone to a variety of weaknesses. They may try to tell their case officer what they think the case officer wants to hear, either out of a desire to please or to increase their monetary value. They often have fixations, even obsessions, that can colour their reporting on personalities or intentions. During the Cold War, the jointly SIS-CIA run Soviet penetration agent Oleg Penkovsky produced impeccable technical information in the form of photographed documents but also political assessments that were so wildly idiosyncratic (often dangerously alarmist) they had to be disseminated in separate serials and under a separate codename from his technical reporting.[4] HUMINT often includes contact with opposition and dissident groups which have their own contacts and sources of information. The problem with sources like this is that such groups have agendas and interests of their own, and they see themselves as collaborators rather than as informants directed and controlled by their handlers in the way individual penetration agents might. Such groups collaborate with an intelligence agency as long as the agency's purpose and their own coincide, but when that ceases to be the case they rarely show any compunction about parting company, or trying to manipulate the agency towards their own end in return. And, of course, human sources are more prone to being used as conduits of deception than technical means. This is as true of entire networks of supposed dissident groups like the Soviet-controlled 'Trust' that purported to be an anti-Bolshevik underground in the 1920s as it is of individual agents such as those played back against their German controllers by the British under the Double Cross programme during the Second World War. As a consequence, the assessors and consumers of intelligence are generally much more cautious about giving credence to HUMINT product than that of technical sources.

In principle, the main overseas collector of HUMINT, the Secret Intelligence Service, has a well-established system designed to evaluate and grade agent product (described by one of the present authors in some detail (Davies, 2000). Attached to each operational Controllerate are Requirements Officers whose task is, as one SIS alumnus has described it, to separate the sheep from the goats amongst a Controllerate's stable of informants, and to ensure that 'sheep remain sheep and goats remain goats' in the appraisal of their product. Case officers can easily get too close to their sources and give them too much credence and trust, not least because a mutual bond of trust is essential to the successful handling of an informant. The Requirements side of SIS, amongst its other tasks, acts as a quality-control mechanism designed to take a detached and objective view of informants and their information.

The so-called '45 minute claim' embodies all of the above problems. It was an isolated source, and if any of the investigative media reporting about the claim is to be believed, it was an officer in the Iraqi military who was in contact with a dissident group in exile called the Iraqi National Congress. The report had apparently come to SIS via the INC. If so, the source for the 45 minute claim represented two very significant problems in terms of the evaluation of their information, and its incorporation into the national assessment process. Firstly, the conduit of the source was a dissident group who had as their prime agenda the overthrow of Saddam Hussein's Ba'athist government. Therefore there would have to be concerns about how SIS' immediate contacts might have been inclined to 'spin up' the confidence or content of the report in order to support their own desire to foment action against Baghdad.

Secondly, and perhaps more disturbingly, the source for the 45 minute claim is an example of a kind of agent called a 'sub-agent' whose information is always more problematic than usual. The use of sub-agent networks is a long-established SIS practice, but one with difficulties in assessing the quality of those networks and their product, difficulties acknowledged by SIS practitioners since the 1930 (see, for example, Whitwell, 1967). In this kind of network a single primary agent contacts and handles an extended network of his/her own sources, and acts as a conduit for their information to SIS via his/her case officer. The advantage of this is that the primary agent can provide SIS with additional sources of information to which the agency would not ordinarily have access. The problem with sub-agents is that the case officer rarely, if ever, has direct, personal access to them to size them up as individuals, debrief them in detail, and develop a clear sense of their access, motivation and reliability. There is even the risk that they may be 'notional', that is, fictitious, invented by the primary agent to look good to his/her handlers or to pad out their financial claims. But even granting the genuine existence of a sub-agent, it is profoundly hard to gauge the reliability of that agent, especially in the long run when one does not have regular access to them.

As a result, the source of the 45 minute claim represented a difficult challenge to evaluators on the Requirements side of the SIS system, let alone to analysts trying to incorporate that information into a more general assessment (about which more below). There was both the problem of skew and spin by SIS primary contacts in the INC and the difficulty of appraising information from an informant to whom SIS operational officers will likely have had little or no access. All of this made will have made it harder to assess whether that the informant was reporting accurately, whether they had been turned, or if they were honestly reporting false information that was being circulated within the Iraqi hierarchy for reasons of morale amongst the lower ranks or currying approval from the upper ranks. To make matters worse, there was no corroborating intelligence to support the 45 minute claim. Although the SIS evaluation of the source was that it was one of long standing and proven reliability, experienced evaluators of SIS product must have known that there would be limits to what inferences and conclusions on could credibly build on that single source.

There appears to have been some SIGINT input. At the time of writing, the main appearance of SIGINT in the available information has been concerned with the investigation of Iraqi efforts to acquire yellowcake uranium ore from Niger. It is worth noting, however, that reference to SIGINT about the Niger case has been limited to fragmentary telephone intercepts serving mainly to prop up more detailed human intelligence (as recounted in the ISC report on Iraqi WMD) as partial corroboration that lends greater credibility to the HUMINT information.

It is probably not sensible to make an accusation of an unsound reliance on HUMINT, or even an reliance on HUMINT from sub-agents associated with dissident groups. This is because there will have been very little intelligence available from other sources. The problem, if any, appears to have rested with the evaluation of that product, and the degree of confidence invested in it.