In the preface to a collection of criminal cases published in 1730, barrister and writer Sollom Emlyn sung the praises of the British legal system. ‘In other countries the Courts of Justice are held in secret; with us publicly and in open view,’ the Irishman wrote. ‘There the witnesses are examin’d in private, and in the Prisoner’s absence; with us they are produced face to face, and deliver their Evidence in open Court.’ Almost three centuries later, the principle described by Emlyn – now known as open justice – remains at the heart of common law legal systems across the globe, including in Australia.

While the rationale for open justice has multiple dimensions, it is grounded in a desire to maintain public confidence in the judiciary. As philosopher Jeremy Bentham observed, and was subsequently endorsed in Scott v Scott [1913] AC 417, the seminal British case on the principle, ‘[p]ublicity is the very soul of justice. It is the keenest spur to exertion and the surest of all guards against improbity. It keeps the judge himself while trying under trial.’ Open justice is better for litigants, better for the judiciary and, ultimately, better for society.

Yet despite the principle’s endurance in the centuries since Emlyn’s writing, it has never been doubted that open justice must permit exceptions. It cannot be absolute. As French CJ noted in Hogan v Hinch (2011) 243 CLR 506 at [20], open justice ‘is a means to an end, and not an end in itself.’ Cases involving vulnerable litigants, or trade secrets, have long been recognised as categories where justice requires secrecy rather than transparency. Justice Rares observed in a recent case: ‘[r]evelation of the secret would destroy its value to the person seeking the court’s protection’: SDCV v Director-General of Security [2021] FCAFC 51 (9 April 2021) at [26].

This tension, between competing public interests in secrecy and transparency, arises most acutely in the national security context. Few doubt that cases involving intelligence information, or that otherwise impact national security, may justifiably require a shroud of secrecy. But such cases can also be squarely in the public interest – because they are often so consequential and arise from a field where there is much executive power and little oversight. When national security is invoked to cloak judicial process, without sufficient scrutiny, confidence in the courts can be diminished.

This post considers two cases where the appropriate balance between open justice and national security has been called into question. In the Witness J matter, a former intelligence officer was imprisoned in secrecy – the individual was charged, arraigned, convicted on a guilty plea, sentenced, and served his sentence, with no public awareness. This state of affairs is currently being reviewed by the Independent National Security Legislation Monitor (INSLM).

In the ongoing proceedings arising from the prosecution of former ACT Attorney-General Bernard Collaery, the National Security Information (Criminal and Civil Proceedings) Act 2004 (Cth) (NSI Act) was invoked to shroud parts of the trial in secrecy. Last year, Mossop J granted the orders sought by the federal Attorney-General under the NSI Act; Collaery appealed, and in early October the ACT Court of Appeal held that at least some parts of the trial must be held in public view. The Collaery case proceeded in parallel with the prosecution of ex-intelligence officer Witness K, Collaery’s client, who pleaded guilty and was recently sentenced.

I will begin by outlining the context to the enactment of the NSI Act. I will then briefly highlight salient features of the Witness J and Collaery cases. I will then proceed with a discussion, drawing on each case, of the constitutional dimensions of open justice in Australia. I will ultimately argue that the NSI Act establishes a scheme with potentially unconstitutional consequences. To avoid these consequences, the NSI Act should be reformed to provide greater safeguards for open justice


The tension between secrecy and open justice in the national security context is not novel. Traditionally, the common law managed the tension through the application of the public interest immunity doctrine. As was said by Gibbs ACJ in Sankey v Whitlam (1978) 142 CLR 1 at [37], ‘the court will not order the production of a document, although relevant and otherwise admissible, if it would be injurious to the public interest to do so’. The doctrine therefore requires a balancing exercise, weighing the relative merit in disclosure and non-disclosure. This was (and remains – both in common law and under uniform evidence law) an exercise for the court. ‘It is in all cases the duty of the court, and not the privilege of the executive government, to decide whether a document will be produced or may be withheld,’ Gibbs ACJ added in Sankey.

The ongoing suitability of this method to appropriately resolve the tension was called into question by the Howard Government following the case of R v Lappas [2001] ACTSC 115 (26 November 2001). Simon Lappas was an intelligence officer who sought to sell classified information to a foreign power. He was prosecuted on several secrecy charges, but – following a successful public interest immunity claim by the Commonwealth in relation to key documents – Gray J permanently stayed one of the charges on the basis that the accused could no longer receive a fair trial.

Against the backdrop of 9/11 and the rise of counter-terror laws and related prosecutions, Lappas raised the prospect that prosecutors might face a vexed choice: reveal sensitive information in open court or be unable to prosecute. (Although in Lappas Gray J was critical of the timing of the Commonwealth’s claim, which was made at a late stage in the proceedings, and suggested the matter might have been handled differently if immunity had been claimed pre-committal: [18]-[20]). Consequently, the Howard Government instructed the Australian Law Reform Commission (ARLC) to review the adequacy of existing measures to protect sensitive information in judicial proceedings and whether they could be improved.

The ALRC ultimately delivered a thoughtful and comprehensive review, of almost 600 pages, in mid-2004. It recommended that legislation be enacted to provide appropriate mechanisms for balancing the tension between secrecy and transparency in national security cases. However, five days before the ALRC was due to report, the Howard Government introduced its own Bill into Parliament aimed at addressing the same issue. This Bill ultimately became the NSI Act. There are a number of significant differences between the ARLC’s proposal and the eventual law, with the ALRC proposing greater judicial discretion and safeguards, while the NSI Act gave greater powers to the Commonwealth Attorney-General.

Witness J

The lack of sufficient safeguards in the NSI Act became apparent in the Witness J case. Witness J, also known by the pseudonym ‘Alan Johns’, was an Australian intelligence officer stationed overseas. While experiencing a mental health crisis, Witness J communicated classified information to his agency via unsecure channels. This, it was alleged, exposed other intelligence officers to risk of harm. Witness J was charged with secrecy offences. He pleaded guilty and in 2019 was sentenced to a term of imprisonment of two years and seven months. He ultimately served 15 months at the Alexander Maconochie Centre, before being released on good behaviour.

None of the above was publicly known until recently. The NSI Act provides a complex scheme for determining the level of secrecy to be placed over proceedings. While the judiciary retains ultimate discretion in reaching this decision, the NSI Act mandates that ‘greatest weight’ must be given to the views of the Attorney-General (ss 31(8) and 38L(8)). However, in practice, this (contested) scheme has proven so technical, time-consuming and no doubt expensive that most NSI Act cases are resolved by agreement between the parties, under s 22. This provides that the Attorney-General, the prosecutor and the defendant ‘may agree to an arrangement about the disclosure, protection, storage, handling or destruction, in the proceeding, of national security information’, and empowers the court to make such orders as considered appropriate to give effect to the agreement.

A s 22 agreement was the cause of the absolute secrecy in the Witness J case, agreed between the parties and given effect by the Judge. If not for a series of fortunate coincidences – reporters at the ACT Supreme Court becoming suspicious about the presence of a security guard outside a closed court room, Witness J filing civil proceedings after his cell was raided and questions being asked in Parliament – we may never have known about Witness J’s plight. Such was the degree of secrecy that even then ACT Minister for Justice Shane Rattenbury, who has oversight of the Alexander Maconochie Centre where Witness J was imprisoned, had no knowledge of the case.


In the Collaery case, the matter has proceeded pursuant to the wider regime established by the NSI Act. The Attorney-General made an application under s 31 of the NSI Act for significant parts of the trial to be conducted behind closed doors. In June 2020, Mossop J made orders ‘substantially in the terms sought by the Attorney-General’: R v Collaery (No 7) [2020] ACTSC 165 (26 June 2020).

Collaery appealed. In early October, the ACT Court of Appeal upheld the appeal: Collaery v The Queen (No 2) [2021] ACTCA 28 (6 October 2021). While a judgment is not yet public, a summary of the reasons of the three-judge panel issued by the Court states:

… the Court doubted that a significant risk of prejudice to national security would materialise [if certain aspects of the case were open to the public]. On the other hand, there was a very real risk of damage to public confidence in the administration of justice if the evidence could not be publicly disclosed. The Court emphasised that the open hearing of criminal trials was important because it deterred political prosecutions, allowed the public to scrutinise the actions of prosecutors, and permitted the public to properly assess the conduct of the accused person.

The Court remitted the matter to Mossop J for further consideration, particularly in relation to the admissibility and effect of further affidavits of the Attorney-General (which Mossop J had not yet considered). ‘Subject to any impact that these affidavits may have,’ the summary of the 6 October judgment notes, ‘there may be public disclosure of information relating to [certain aspects of the trial].’

It is expected, given the public interest in the case and the relative positions of the parties, that the matter will be heard by the High Court – either arising shortly with an appeal by the Commonwealth Attorney-General from this ACT Court of Appeal decision, or via another appeal to the Court of Appeal (and then ultimately to the High Court) following reconsideration by Mossop J.

Notably, both the hearing before Mossop J, and the Court of Appeal, were in closed court. As Mossop J explained at [3], ‘[t]hat is because the “closed hearing requirements” set out in s 29 of the NSI Act applied to the hearing: NSI Act s 27(5). This meant that members of the public and the media were excluded from the hearing.’ Any eventual High Court appeal would be heard in closed court (although, presumably, the special leave application would not, given it is not, strictly, a s 29 hearing for the purposes of the NSI Act).


I now wish to explore three distinct but related observations that follow from the above. The first point is that the constitutional status of open justice remains underdeveloped, and that lacuna is problematic when the tension between secrecy and transparency arises in cases such as these. The second is that there should be a constitutional minimum standard of open justice that cannot be subverted, such that fully secret prosecutions – like that in Witness J – are unconstitutional. The third related point is that the opacity of the current Collaery proceedings might risk undermining public confidence in the judiciary in ways that should be contrary to the Constitution. It follows from these points that reform of the NSI Act is desirable.

The Incomplete Constitutionalisation of Open Justice

My first contention is that the recognition of open justice as a constitutional principle remains incomplete. Following the seminal British open justice case of Scott v Scott, it took less than a year for the case to be endorsed in Australia, with Barton ACJ holding in Dickason v Dickason that ‘one of the normal attributes of a Court is publicity, that is, the admission of the public to attend the proceedings’: (1913) 17 CLR 50, 51. In 1976, in Russell v Russell (1976) 134 CLR 495, the High Court invalidated a federal law that purported to require state courts to hear family law matters behind closed doors, albeit on the basis that the Commonwealth lacked legislative authority, rather than because of a distinct constitutional limit.

In subsequent decades, dozens of High Court cases have underscored the centrality of open justice to the exercise of judicial power. In a remarkable case in the 1980s, Deane J issued a heated statement after intelligence officers had sought to take the names of members of the public attending the High Court. He observed:

As a general rule the Court’s exercise of that judicial power is in public sittings … One reason for that approach to the exercise of judicial power is that open and public administration of justice by the country’s final Court is a safeguard of judicial independence and conducive to public trust: ‘Orders in ASIS Secrets Case’ (1988) 19 Legal Reporter 6.

In Grollo v Palmer (1995) 184 CLR 348, McHugh J said at [29]: ‘[o]pen justice is the hallmark of the common law system of justice and is an essential characteristic of the exercise of federal judicial power.’

In the post-Kable era, a number of cases have touched on the constitutional salience of open justice. None have been successful on that ground: see, eg, Gypsy Jokers Motorcycle Club Inc v Commissioner of Police (WA) (2008) 234 CLR 532; Hogan v Hinch (2011) 243 CLR 506; Condon v Pompano Pty Ltd (2013) 252 CLR 38. Thus, Professor James Stellios summarises in the 2nd edition of his text The Federal Judicature: Chapter III of the Constitution, at 555:

while there may be a general rule that judicial proceedings shall be conducted in public – a rule that is now constitutionalised as a feature of State courts – that rule is not absolute … The accommodation of the constitutional value of open justice and competing interests will no doubt require further refinement and calibration.

What is noteworthy about the post-Kable jurisprudence is that, despite the regular proclamations about the importance of open justice, the principle has been largely developed as a component of natural justice and procedural fairness. A number of the post-Kable cases have involved ex-parte hearings or secret evidence – such that the focus of the invalidity inquiry was on unfairness to the other party, rather than open justice to the world at large. As a result, open justice as a standalone constitutional value remains underexplored. This is problematic in cases such as Witness J and Collaery because of the lack of clarity about the extent to which the value will bend to competing interests (especially in Witness J, where the parties had agreed on the secrecy, such that the natural justice dimension was moot). The status quo makes it difficult, if not impossible, for organisations seeking to vindicate open justice – such as media outlets – to do so.

The Necessity of Constitutional Minimum Standards

Notwithstanding this uncertainty, I think it follows from the clear acceptance of open justice as a constitutional value that there should be a minimum open justice floor below which a court cannot go – in other words, there are necessary limits on legislative and judicial power in requiring, permitting or undertaking secret hearings. On this view, a Commonwealth law could not validly require or authorise the making of an agreement that has the effect of a fully-secret hearing, nor can a judge order it (whether with legislative authorisation or otherwise). I am not arguing that no level of secrecy is permitted, but merely that fully secret hearings are unacceptable in our constitutional system.

Imagine a federal law that directed the Federal Court to hear all matters involving defamation claims brought by serving Ministers in complete secrecy. Such a law would be repugnant to the exercise of judicial power and undermine public confidence in the court. Such a law, surely, would be unconstitutional. If, instead, there was no such law but a plaintiff sought, by application upon commencement, for their case to be heard in total secret, I would submit that it is beyond judicial power to grant such an order.

The necessary duality of this limitation is underscored by the Witness J case, where the secrecy arose via agreement of the parties, given effect by the ACT Supreme Court via s 22 of the NSI Act. While it might be said that the Court should not have accepted the s 22 agreement (and, indeed, the NSI Act does not require, only provides, that the Court can make orders giving effect to an agreement), ultimately the order was made. Accordingly, the question of limits must consider both legislative and judicial authority. It might be remembered that the foundational British case, Scott v Scott, focused predominantly on judicial power rather than legislative authorisation.

The constitutional need for such minimum standards is heightened by the recognition that other parties have a legitimate interest in open justice – and, indeed, standing to challenge exceptions to open justice in particular cases. It has been accepted, albeit not universally, that the media has standing to intervene in cases involving open justice – to challenge suppression orders, for example (although this area is not without complexity: see Michael Douglas, ‘The Media’s Standing to Challenge Departures from Open Justice’ (2016) 37 Adelaide Law Review 69). This standing – the recognition that there is a sufficiently significant interest held by a third party to permit their involvement in proceedings – is a nullity if a case is heard in total secret because there would be no knowledge of the case and hence no ability to contest the secrecy.

This may seem an obvious point, but I consider it to have compelling relevance. Even if it was accepted that there might be extraordinary circumstances were the complete denial of open justice was desirable, cloaking a case in total secrecy – as in Witness J – permits no safeguard to the possibility that the balancing act was wrongly decided. It is impossible to contest or appeal a decision that no-one knows about. It must follow from this recognition that open justice is a principle of significance to more than just the parties to a case that the de facto exclusion of review of a decision to hold a case in total secrecy is impermissible. On the basis of what occurred in Witness J, there is nothing to say that another secret trial is not happening in the Australian judicial system right now.

Against this argument, it might be put that if natural justice can be reduced to nothingness, so too can open justice in the appropriate circumstances. In Leghaei v Director General of Security [2005] FCA 1576 (10 November 2005), the applicant’s visa was cancelled on national security grounds. He sought judicial review on the basis that no allegations had been put to him and hence he had been denied natural justice. At first instance, it was held that ‘the potential prejudice to the interests of national security involved in such disclosure appears to be such that the content of procedural fairness is reduced, in practical terms, to nothingness’: at [88]. This holding was not disturbed on appeal, in a heavily-redacted judgment: Leghaei v Director-General of Security [2007] FCAFC 37 (23 March 2007).

However, the critical distinction is that, in Leghaei, the applicant still knew the adverse decision had been made. He could therefore exercise his constitutionally-entrenched ability to seek judicial review (even if that was to prove a fruitless exercise). In contrast, in a Witness J scenario, other interested parties (such as the media) have no ability to contest the appropriateness of the secrecy imposed on a case. It may well be that, in extraordinary circumstances, the minimum standards guarantee only minimal disclosure of the bare fact of a case taking place. That may not be much. But it is better than nothing – and enables interested parties to seek review by a superior court.

Why Collaery Should Be Unconstitutional

Finally, I want to focus on the anticipated High Court appeal in the Collaery matter. While the Court of Appeal’s decision is a welcome development, and underscores the importance of open justice, I suspect its judgment is not the end of the road in this matter. Further litigation around the NSI Act raises the spectre, by operation of s 29, that the High Court will be required to hear a case of upmost national importance behind closed doors. This is particularly repugnant because s 29 leaves the court no discretion to appropriately manage secrecy concerns – it is a mandatory requirement – and because, by the time the case reaches the final appellate level, it will mostly involve arguments about law, rather than potentially sensitive facts. Even if it was accepted that part of the High Court hearing might justifiably be held in closed court, requiring that the totality of the appeal be heard in such a manner risks undermining public confidence in the Court.

The constitutional validity of s 29 has been considered once before, in R v Lodhi [2006] NSWSC 571 (7 February 2006). In that case, a prosecution under anti-terror law, the defendant and media companies challenged the operation of the NSI Act. The challenge was unsuccessful and failed on appeal (special leave to appeal to the High Court was, in turn, denied). The case focused predominantly on Chapter III concerns with the ‘greatest weight’ requirement in the ultimate secrecy decision, but the media had also challenged the s 29 mandatory closed court (on implied freedom of political communication grounds). Justice Whealy rejected these arguments, largely because he characterised a s 29 hearing as a narrow, procedural matter:

[i]n my opinion, the fact that the … hearing is to be a closed hearing does not place an undue burden given the legitimate aim of such a hearing and the subject matter with which it deals. It is a limited hearing dealing with a limited topic as I have indicated: at [123].

Lodhi can therefore be distinguished from Collaery. It cannot be said that a High Court hearing in such a significant, politically-charged case is a ‘limited hearing dealing with a limited topic’. As and when the Collaery proceedings reach the High Court, the operation of the NSI Act will risk negatively impacting public perceptions of the Court. Unfortunately, due to the absence of clarity around open justice as a standalone Chapter III value, and the frail nature of the implied freedom, it is far from clear that a constitutional challenge to s 29 in the context of Collaery would succeed. That is a great pity.

The obvious response to these shortcomings is legislative reform. The NSI Act was introduced hurriedly, and ignored key aspects of the ALRC report that was intended to underpin it. The INSLM will soon finalise his inquiry into the Witness J case – his ultimate recommendations may well provide a useful opportunity for wider NSI Act reform. Greater safeguards for open justice are essential – including a prohibition on entirely secret cases (as in Witness J) and the allowance of judicial discretion in relation to closed s 29 hearings.


In Secrets and Leaks: The Dilemma of State Secrecy, a thought-provoking 2013 text, Rahul Sagar argues that secrecy is both good and bad for democracy. The tension, he argues, cannot be resolved but must instead be managed through sensible institutional design and individual action (he writes in favour of whistleblowing and leaking, in appropriate circumstances). Sagar’s contribution is valuable because it prompts us to think not of a binary – secrecy or transparency – but of a continuum – circumstances where secrecy is necessary and justifiable, circumstances where it is evidently not, and difficult situations in between.

The issue underscored by the Witness J and Collaery cases is that, when secrecy is invoked improperly (at one end of the spectrum), it undermines a system that otherwise manages to retain public confidence despite its anti-democratic nature (at the other end, or in the middle). In Witness J, for example, significant information about the prosecution has subsequently been released to the INSLM and made public, with the Commonwealth’s consent. This indicates that the blanket secrecy at the time was unnecessary – at least some information could have been made public contemporaneously. In the Collaery case, meanwhile, it is widely accepted as fact that Australia spied on Timor-Leste. The attempted use of the NSI Act in the case is not to keep that detail public, but to enable the Commonwealth to admit it in court (to enable effective prosecution of the accused) while maintaining the public position of ‘neither confirm nor deny’. (I am not exaggerating this intentional deception – Mossop J states as much at [10] in his judgment).

The problem with the ‘trust us’ model of executive and judicial government in the secrecy context is that, when the public is given a valid reason to distrust, confidence is eroded even for cases where trust is justified. The legitimacy of our approach to the entire spectrum of secrecy is brought into doubt. Open justice has served us well for centuries – and, as I have argued, has constitutional salience in Australia. We undermine it at our peril.

Kieran Pender is a Senior Lawyer with the Human Rights Law Centre. With thanks to Dr Rebecca Ananian-Welsh and Daniel Webb for their input.

Suggested citation: Kieran Pender, ‘Witnesses J, K – and L? Open Justice, the NSI Act and the Constitution‘, on AUSPUBLAW (12 October 2021) <>(opens in a new t