Threading the Needle: Structural Reform & Canada's Intelligence-to-Evidence Dilemma

Becuase I am a patriot, and wasn't available to sell my country out today in Finland, I have written yet another paper on intelligence-to-evidence. This one tries to straddle the distance between "accessible for non-lawyers" and "technical enough for lawyers". I try hard in this paper to lay out what intelligence-to-evidence is, in my view. Most importantly, I propose what I call "moneyball" solutions to this problem, expanding and refining those I have suggested elsewhere and supplementing the solutions that have been raised by others (which as mostly complementary). I have spent a lot of time talking to people about this, and nothing I have heard has persuaded me things can't be done better. It is not quite a Gordian a knot as many seem to assume. On the other hand, there is no "home run" solution. A lot of players will need to come to the table with renewed determination. The paper is intended as a draft working paper. I welcome comments and feedback. It may be downloaded here.

The paper's abstract is as follows:

This article canvasses the “intelligence-to-evidence” dilemma in Canadian anti-terrorism. It reviews the concept of “evidence”, “intelligence” and “intelligence-to-evidence” (I2E). It points to the legal context in which I2E arises in Canada. Specifically, it examines Canadian rules around disclosure to the defence: the Stinchcombe and O’Connor standards and the related issues of Garofoli challenges. With a focus on CSIS/police relations, the article discusses the consequences of an unwieldy I2E system, using the device of a hypothetical terrorism investigation. It concludes disclosure risk for CSIS in an anti-terrorism investigation can be managed, in a manner that threads the needle between fair trials, legitimate confidentiality concerns and public safety. This management system rests on three legs:

  • Manage the relevance “tear-line” so that crimes less intrusive on CSIS information holdings are preferred over ones that are more intrusive. This strategy requires applying a prosecutorial insight to those investigations and planning their conduct to not prejudice trials. I bundle this concept within the category of “collecting to evidential standards” and “managing witnesses”.
  • Legislate standards to create certainty from the murk of evidence law. Here, two innovations stand out: legislate O’Connor style third-party status for CSIS where: CSIS’s investigation is a bona fide security intelligence investigation; CSIS and police do not have full, unmediated access to each other’s files; and, CSIS does not take an active role in the police investigation. But do not build this legislated third-party status around rigid barriers on information-sharing. Second, legislate ex parte, in camera procedures for Garofoli challenges of CSIS warrants in which special advocates are substituted for public defence counsel.
  • Manage the public safety risk by creating a fusion centre able to receive investigative information from all-of-government and fully apprised of the public safety risks associated with an ongoing investigation (or parallel investigations). Ensure it includes representatives from all the services with legal powers to respond to threats. The fusion centre would not itself be an investigative body, and would have O’Connor-style third-party status, something that would not require legislation but which might benefit from it.

Staying Left of the Bang: Fixing Canada's Dysfunctional System of Parallel CSIS/RCMP Anti-terror Investigations

Canada has what is known as an “intelligence-to-evidence” problem.  The first step is recognizing it: We do not do a good job deploying information collected by our intelligence services as evidence that can be used in anti-terrorism prosecutions.  This is probably one reason why we have many fewer terrorism prosecutions than seem warranted by our threat environment – and a tendency to resort to other, often imperfect measures (like immigration tools and peace bonds).

More than this, an exaggerated fear that sensitive intelligence will be torn from it and disclosed in criminal proceedings makes CSIS very reluctant to cooperate fully with police. The result is a peculiarly awkward Canadian approach to anti-terrorism investigations condemned starkly by the Air India bombing inquiry.  This is a system that is suboptimal at best and downright dangerous at worst.

In a first post on this issue, I examined how the United Kingdom has addressed the issue of intelligence-to-evidence, in a manner much more sophisticated than anything echoed in Canada.  But received wisdom in Canada seems to be that we cannot emulate the United Kingdom, because of law. And specifically, fingers are often pointed at constitutionally-mandatory rules on disclosure to the defence in criminal proceedings.

In my view, this is an example of law being blamed to justify cultural inertia in our security services.  The real risk is not that CSIS information will be disclosed, involuntarily.  The real risk is that CSIS does not manage its affairs so that information it collects is useful for an all-of-government anti-terrorism effort. 

I lay out my reasons for this opinion in this blog entry.


A. Disclosure Obligations 101

Stinchcombe Standard

The focus in this blog is on criminal law disclosure. This is the area where the state’s disclosure obligations are most robust.[1] Here, the starting point is the Supreme Court of Canada’s famous (or infamous, depending on your perspective) Stinchcombe decision. Pointing to section 7 of the Charter, the Court found a general duty on the Crown to disclose all relevant material to the defence, even where it intends to make no reliance on this information. Nothing turns on whether this information is exculpatory or inculpatory: the threshold standard is one of “relevance”. And so, among other things, the Stinchcombe standard entitles the defence to unearth and see government affidavits used to support wiretap applications.


Limits on Stinchcombe

1. Relevance Standard

In practice, Canada’s criminal law disclosure standard is broader than that in the United States or United Kingdom – in consequence, this is often where an argument that we cannot emulate these jurisdictions on intelligence-to-evidence begins and ends.

But there are limits on Stinchcombe. For one thing, the Crown has no obligation to disclose information that is “clearly irrelevant”.

2. Control Standard

Even more importantly, the disclosure obligation is on the Crown – in practice, that means prosecutors and the police in criminal matters. It does not extend to all government agencies – so-called third parties. And so CSIS has been treated as a third party, at least so long as its investigation is not so interwoven with that of the police that courts regard the two as conflated.

This does not mean that a government third party (in this case, CSIS) has no disclosure obligations. But it does mean that those disclosure obligations are different, governed by what is known as the O’Connor standard. That standard sets a higher threshold on disclosure than does Stinchcombe. So, for instance, the accused must persuade a trial judge that “there is a reasonable possibility that the information is logically probative to an issue at trial or the competence of a witness to testify” (known as the “likely relevance” standard). If the accused succeeds, then the judge will order production of the information for the judge’s own review. And then, in a second stage, the judge weighs the different considerations favouring disclosure or non-disclosure to the accused. Exactly what considerations would go into this calculus in a CSIS case are unclear: there is no legislative guidance here, as there have been in other circumstances.

3. Privileges

Moreover, Stinchcombe does not annul privileges in the law of evidence, including police informer identity privilege. That privilege persists, although it too has an outer limit: it does not apply to identity information that goes to the very question of innocence or guilt. This “innocence at stake” qualifier exists where there is “a basis on the evidence for concluding that disclosure of the informer’s identity is necessary to demonstrate the innocence of the accused”.

Critically, there is also a national security privilege codified in section 38 of the Canada Evidence Act. It is important to understand how this privilege works. Under section 38, specially designated Federal Court judges decide whether material, if disclosed to the accused, would harm national security, national defence, or international relations. If it would, judges then balance this injury against the fair trial virtues of disclosure, and may protect the information from disclosure or may order the information disclosed (albeit perhaps in redacted or summarized form).

The section 38 process can be unwieldy. The disclosure decisions made by the Federal Court are generally made before the terrorism trial starts, and the process can be long and fraught. Following this disclosure litigation in Federal Court, the trial judge then has to accept whatever non-disclosure decision the Federal Court makes. But: the trial judge also must then make the difficult decision whether to halt the prosecution because the Federal Court’s non-disclosure order has made the trial unfair. This also is a difficult proposition. The trial judge may not even know the specifics of the secret information.

A defendant in the Toronto 18 case challenged this whole system on constitutional grounds. The Supreme Court of Canada recognized that the two-court section 38 system could “cause delays and pose serious challenges to the fair and expeditious trial of an accused, especially when the trial is by jury,” but decided that it was constitutional because the trial judge could always stop a trial, should the Federal Court’s non-disclosure order make it impossible for the accused to have a fair trial. The Court stressed that “the trial judge may have no choice but to enter a stay.”

Some participants in the case argued that this approach “puts the Attorney General and the trial courts in the dilemma of playing constitutional chicken.” For its part, the Court expressed the hope that a sensible application of section 38 would avoid such a result, perhaps using the intermediary of a security-cleared special advocate as a link between Federal and trial courts. But bottom line: this is a taxing system that greatly complicates many Canadian terrorism trials.

It is very different from the “single-court” UK and US approach, where the trial judge him- or herself decides whether the information should be protected by a national security privilege.

The Canada Evidence Act has another critical feature: the Attorney-General’s certificate.  I shall discuss that below.


B. What CSIS Fears

Disclosure of “Crown Jewels”

CSIS fears disclosure of its information, methods and sources in open court, and especially fears being subjected to Stinchcombe disclosure. In an undated memo on this topic, it writes:

Despite mechanisms to protect sensitive information, litigation involving issues of national security often results in lengthy legal processes that are resource intensive and that jeopardize national security through the production and disclosure of sensitive information.  Beyond the considerable financial resources required to support these cases, the production and disclosure of sensitive information also entails a substantial risk to national security. While … provisions provide for the protection of classified information in s.38 CEA proceedings, the Court may decide that disclosure is required if the public interest outweighs its protection. As a result, numerous disclosures have and continue to be made concerning CSIS investigative interests, tradecraft, human SECRET source operations and information provided by foreign partners. These disclosures negatively impact CSIS operations and adversely affect Canada's national security interests. 

Managing Fear

To manage this disclosure risk, CSIS structures its relationship with the police to take advantage of the Stinchcombe/O’Connor distinction.  That is, it maintains parallel, rather than conjoined investigations, so that it is not conflated with the police for purposes of disclosure.  As CSIS asserts in the above-noted memo,

to mitigate such issues [of disclosure], the Service has, with its partners, created a whole range of processes to protect classified information that may be used to inform enforcement actions. For instance, the One Vision framework for cooperation between CSIS and the Royal Canadian Mounted Police (RCMP) was enhanced to avoid inadvertent disclosure of CSIS information to RCMP. This, given that information shared with the RCMP may appear in their case files, judicial authorisations and disclosure packages to the Court as part of criminal prosecutions, ultimately becoming subject to disclosure obligations.  

But this approach creates artificial bureaucratic siloes between agencies often investigating the same target.  Such siloes – known as “less is more” information-sharing -- are dangerous (and possibly disastrous) in anti-terrorism.  There are now many instances in which CSIS has failed to share information about dangerous people with the RCMP. (See here for a recent example). As Kent Roach and I write in False Security:

Most Canadians would be — and should be — both shocked and scared by revelations that CSIS does not share all of its intelligence about terrorist conduct with the police, even when they know the police are looking in the wrong places. From the outside, “less is more” looks crazy and dangerous.  Even with their amped up Bill C-51 powers and with the best intelligence in the world, CSIS cannot arrest a person for criminal law purposes.  Nor can it lay charges, leading to a prosecution and incarceration.

The most galvanizing recent statement on the dangers of “less is more” was offered to the Senate security committee studying Bill C-51. Joseph Fogarty, a former UK government security liaison in Canada, condemned Canada’s arrangement, noting how poorly it compared to the much more seamlessly coordinated anti-terror endeavours of the UK police and the MI5. Commenting on whether he thought the United Kingdom has anything to learn from Canada’s operational arrangement between CSIS and the RCMP, Fogarty observed: “with this particular regime, with the greatest of respect, I wouldn't incorporate a single aspect of it, at the minute, because it’s dangerous.” And averting to CSIS’s failure in the Toronto 18 matter to alert the RCMP to the terrorist training camp and its errant surveillance operation, Fogarty stated:

if you take that sort of decision on an operation and are running up against fast-moving, sophisticated opponents, the consequences could be a tragedy. . . . [I]magine a situation in which an MI5 team discovers a Provisional IRA camp in the U.K. and decides not to tell anybody about it. . . . It’s a tragedy waiting to happen. You have been remarkably lucky, as a country, that you have not faced fast-moving, sophisticated opponents since 2001 because you could have been living in tragedy here.

We will not always be lucky.

In truth, thoughtful people in government acknowledge that the present circumstances are unacceptable, and that we live on borrowed time.  But many in government seem to think that we need to simply live with this situation; that the disclosure/security trade-off is an impossible dilemma that cannot be solved, because of the Supreme Court’s extreme approach in Stinchcombe.


C. CSIS Exaggerates Legal Risk

CSIS has, however, exaggerated legal risk: its intelligence is not at serious risk of disclosure if it manages its affairs properly.

Attorney General’s Certificate

First, the government can stop information from being disclosed using section 38. Period.

In the memo above, CSIS suggests that Federal Court disclosure orders under section 38 have been prejudicial to security.  This is a sweeping statement, unsupported by any details and easy to make where the Service is unwilling to justify its conclusions. 

But even if it is true, the conclusion must be that the Service has simply decided to allow this information to be released. That is because the government has, essentially, an absolute ability to stop disclosure under section 38, using what is known as an “Attorney-General’s certificate”.

This certificate allows the government to short-circuit a court disclosure order. Section 38.13 of the Act empowers the Attorney General to personally issue a certificate “in connection with a proceeding for the purpose of protecting information obtained in confidence from, or in relation to, a foreign entity as defined in subsection 2(1) of the Security of Information Act or for the purpose of protecting national defence or national security.” Issuance of the certificate has the effect of barring any subsequent disclosure of the information in a proceeding for ten years (and for a further period if the certificate is renewed at the end of that ten years). In other words, the certificate may reverse an order from the Federal Court authorizing disclosure under section 38, subject to a very narrow and limited appeal before a single judge of the Federal Court of Appeal.

As best I know, the AG certificate has never been used since the creation of this power in 2001.  It is almost as if the government has forgotten it exists: it goes unmentioned in the various government documents discussing the problem of protecting national security information that I have reviewed.

But bottom line, given the scope of the AG certificate: if information is disclosed that prejudices national security, it is the government’s fault. As Justice Canada counsel Don Piragoff told the Senate when the provision was enacted: “The provision is a last resort for the Attorney General to ensure that information critical to national security is not disclosed in judicial proceedings to which the Canada Evidence Act applies or through other government processes. … The certificate issued by the Attorney General … would be the ultimate guarantee that information such as sources of information and names of informers would not be made public. …”

Legislated CSIS Source Protection

The presence of the “nuclear” AG certificate is alone an answer to the CSIS concern that its sensitive information risks being extracted from it, against its firm resistance.  But it is also worth noting the amendments made in 2015 codifying a special CSIS source identity protection in section 18.1 of the CSIS Act itself.  CSIS now has a statutory privilege protecting its source’s identity as potent as that existing for police informants.


D. The Real Issue: CSIS Cultural Baggage and System of Operating

So to clear: the government can protect CSIS’s information from disclosure.  That compelled disclosure issue is a red herring, and yet is the animating impulse behind our awkward “less is more”, dysfunctional organization of CSIS/RCMP anti-terror investigations.

This then begs the question: what is the true concern?  In my view, the real issue is not that secret information will be yanked out of the Service.  Instead, it is the fact that CSIS secrets are not useful.  Specifically, CSIS sensitivity to disclosure of what it has called its “Crown jewels” means that if it does share, police and prosecutors will need to suffer through protracted Canada Evidence Act proceedings (or other disclosure disputes) to protect the sensitive bits.  This risk scuttling prosecutions, either out of delay or because a trial judge ultimately concludes that some form of cheery-picked CSIS disclosure prejudices a fair trial.

Put another way, the problem is not that CSIS can’t protect its secrets.  The problem is that no one can use those secrets to put bad guys in jail without risking a trainwreck in court.  Or, stated differently: The problem of intelligence to evidence is not “we can’t protect our really important secrets”.  Instead, the problem is that CSIS does not collect information in a manner useful to an all-of-government anti-terrorism strategy.

The key issue is, therefore, whether CSIS can collect information (and also, now do threat reduction) in a manner that is more useful.  As I describe at length in discussing the UK experience, it is clear than a CSIS’s closest counterpart -- MI5 -- is perfectly capable of collecting intelligence to evidential standards that then can be re-deployed for prosecutorial purposes.

Why has Canada not done this?  Exaggerated fear about Stinchcombe and the myth of involuntary disclosure is one reason: if you believe your information will be extracted if you get too close to the police, then you have little incentive to think creatively about reforming that relationship.

But as I have suggested, this is a myth: The combined effect of legislated source identity protection, Canada Evidence Act balancing and then, if all else fails, the nuclear option of an AG certificate means that you do not need to fight tooth and nail to remain in O’Connor space and avoid Stinchcombe

And once you appreciate that the risk of involuntary disclosure can be controlled, the emphasis shifts to a more productive discussion: how can CSIS change its modus operandi to create “clean” information that is not so thoroughly and irremediably intertwined with truly sensitive information, means, methods and sources? In other words, how can CSIS be more useful to criminal justice outcomes – ultimately, the most potent remedy the state has in anti-terrorism.

The UK experience (discussed in my prior blog post) points the way: close, synergistic cooperation, co-location; and protocols on collection to evidential standards.  This will require a cultural and operational shift at CSIS (and, no doubt, with the police).  For example, CSIS will need to align its practices (in anti-terrorism at least) with a weather eye on downstream implications for prosecutions.  For instance, it may not operate sources and agents and officers in a manner that amounts to entrapment.  (But even now, if it does that, and the defence is able to find out, the case will collapse – so if it does operate in this manner, we already have serious rule of law issues, just ones that are masked by secrecy. This translates into “scandals in the making”.)

Resistance to such changes would likely be fierce, institutions being what they are.  But the alternative is persisting with a system of parallel investigations that, from a security perspective, is irrational – and then waiting for the next avertible security disaster, and a re-run of the Air India inquiry condemning the way we run anti-terrorism in Canada.


E. Concrete Steps

Let me end this essay with five thoughts on concrete steps that might propel change:

  • Like the Air India bombing inquiry, I do not believe that CSIS should have the final say on whether its intelligence should be prioritized for intelligence or evidential purposes. Seamless integration, of the sort practices in the UK, places the focus on public safety, and establishes an inter-agency process for deciding how information should be used.  We need a true inter-agency process, not a bureaucratic One Vision process in which CSIS is de facto in the driver seat and other players do not have full access to information.
  • Legislate a standard for CSIS disclosure that codifies the Stinchcombe obligation in the specific instance of intelligence.  I am not proposing an effort to legislatively roll-back Stinchcombe – a constitutional norm.  Instead, I believe that certainty as to how Stinchcombe applies in the specific area of intelligence and anti-terrorism would provide predictability that could fuel the more seamless relationship between police and CSIS so essential to modern anti-terrorism.  In some large measure, legislative codification aided reform in this area in the United Kingdom. 
  • And: eliminate the bifurcated court system for section 38 proceedings.  It is not rocket science to come up with a system in which trial judges specialized in terrorism cases handle the trial and the disclosure issues.
  • Audit CSIS investigative practices to assess exactly what collection to evidential standards would mean in terms of reforming CSIS practices.  In truth, CSIS must have a least started reconsidering its intelligence methodologies, given the expectations the Supreme Court has already imposed on it as part of the Charkaoui II standards.
  • Once all of this is cleaned up, and there are still instances where CSIS cooperation with police in anti-terrorism risks revelation of secrets of real sensitivity, do not revert to a counterproductive effort to stay in O’Connor space by keeping police at arm’s length.  Rather, maintain the close relationship and then protect the secret using the Canada Evidence Act, including (if necessary) an AG certificate. It may mean that some cases cannot be prosecuted – but I suspect that such cases will be less frequent than under present circumstances.


[1]           An argument can be made that closed-court disclosure to special advocates in immigration security certificate proceedings may be even more sweeping, although that argument may not survive the changes made to that regime by Bill C-51. The latter rolled back the sweep of at least some of those disclosure procedures.

New "RCMP Accountability, But... Act" coming into Force

Public Safety Minister Blaney announced today the coming into force of 2013's Enhancing Royal Canadian Mounted Police Accountability Act.  A substantial redrafting of the RCMP Act, the accountability measures in the amendments are an improvement, but come with important catches in the area of national security.  So, tongue in cheek, I have called this blog entry "RCMP Accountability, But... Act".

Readers will recall that the Arar Commission -- rapidly disappearing behind the taillights of inaction -- recommended a robust review body for the RCMP, with powers analogous to those of CSIS's Security Intelligence Review Committee (SIRC).  The key recommendations were an enhanced arm's length complaints (and also review) body which could a) self-initiate reviews on national security matters; b) investigate and report on complaints; c) conduct joint reviews with SIRC and the CSE commissioner; d) conduct reviews at the request of the public safety minister; e) conduct reviews into other government bodies on request of the governor in council (the new body was to extend beyond just the RCMP to include other, currently unreviewable bodies like CBSA); (f) make recommendations to ministers in the course of all these activities.  More than this, the body was to have (g) robust access to the information required for its job.

This is report card season.  And so measured against these recommendations, the new Civilian Review and Complaints Commission for the RCMP gets a D, mostly because of the information access nightmare it sets in  play and also because the government continues to resist the idea that accountability should be streamlined and coordinated, not stovepiped by agency.   Chart 1 gives the details:

Arar Recommendation

Actual Power

a) self-initiate reviews

B grade.  The power exists, but it is not as regularized as the equivalent review function of SIRC and is caveated with provisos about only doing this if there are sufficient resources.  In other words, this is painted as an extraordinary event, not a regular undertaking.

b) investigate and report on complaints;

A grade.  Looks fairly good, subject to concerns about information access in (g) below.

c) conduct joint reviews with SIRC and the CSE commissioner;

F grade.  While there is a language about collaborating with other police review entities, nothing about working with the national security review bodies.  In other words, the new Act misses the entire point of the Arar recommendations.  (For how I would propose fixing it, see the back end of this post.)

d) conduct reviews at the request of the public safety minister;

A grade.  Looks fairly decent, subject to concerns about information access in (g) below.

e) conduct reviews into other government bodies on request of the governor in council (the new body was to extend beyond just the RCMP to include other, currently unreviewable bodies like CBSA);

F grade.  The Commission’s mandate is limited to the RCMP.  Other powerful agencies – not least the CBSA – sail about still without any review body at all.  Very bad news.

f) make recommendations to ministers in the course of all these activities. 

A grade.  Where there is a review, recommendations can be made.

g) full access to information

D- grade.  Herein lies the rub.  The Arar commission envisaged the Commission having SIRC-like access to information (that is, basically unalloyed, at least in principle).  The Act improves on its predecessor, surely, but in the realm of so-called “privileged” information (including that most material to national security matters), it allows the RCMP to resist disclosure on the basis that the information is irrelevant or unnecessary, whatever that means (in the eyes of the Commissioner).  This then provokes a complicated adjudication process by a retired judge.  And after seven years, an endless administrative process, several judicial review applications, and millions of dollars expended in Canada v. Canada cases, maybe the Commission gets the information and the review/investigation can continue.  Or maybe not.  This is not the SIRC model.  At best it is SIRC-lite, or maybe even SIRC-less. 


Because of the shortcomings in the information access area, every other improvement in the system risks being illusory. 

A last observation stems from another Arar commission recommendation: "Appointments of [Commission] members should be aimed at inspiring public confidence and trust in their judgment and experience. Appointees should be highly-regarded individuals with a stature similar to SIRC appointees."  Strong agencies with weak appointees are only as strong as their appointees.  Weak agencies with strong appointees can be stronger than their legislative framework.  Ten years post-Arar, SIRC is not a model for uncontroversial and confidence-inspiring appointees, and an even weaker model for proper staffing and financing levels.  Overall, the record of watchdog appointees over the life of the current government is checkered: some very good, some horrific. 

So the final question: who will these commissioners be, and will they be adequately resourced? Unless they are rock-solid, and well-financed and staffed, then the already structurally-imperfect Enhancing Royal Canadian Mounted Police Accountability Act will be empty legal formalism.