In this issue


Recent trends in litigation – the rise of the class action against the Commonwealth


Increasing your chances of successfully calling on an unconditional bank security


Technology assisted review (TAR


Disqualified senators

Class actions

Recent trends in litigation – the rise of the class action against the Commonwealth

In recent years, AGS Dispute Resolution has seen an increase in the number of representative actions1 brought against the Commonwealth. These actions raise unique and significant legal issues which the Commonwealth, and those representing it, must consider. Due to the size and complexity of the matters, class actions require collaboration with multiple government agencies and departments and a pragmatic approach to the resolution of the issues in dispute.



Marcella Bienvenue
Marianne Peterswald

Marcella Bienvenue
Senior Lawyer
AGS Dispute Resolution
T 08 8205 4261


Marianne Peterswald
Senior Executive Lawyer
AGS Dispute Resolution
T 02 6253 7260


The rise of the class and representative action

The types of matters which have recently been commenced as representative actions against the Commonwealth arise out of a myriad of different factual situations and out of different legislative and common law causes of action. They have been brought in both State Supreme Courts2 and also in the Federal Court of Australia.

The breadth and scope of the types of matters which have recently been litigated and resolved, as class actions against the Commonwealth, and in which AGS has acted, include the following matters:

  • Duval-Comrie v Commonwealth: In 2016, the Federal Court approved the settlement of a complex and long-running class action brought on behalf of over 10,000 intellectually disabled members against the Commonwealth. In that case the class members alleged discrimination by the Commonwealth in developing and promoting the use of the Business Services Wage Assessment Tool (BSWAT) – a tool used to assess the wages of workers with disabilities in supported employment.

The settlement involved legislative amendment to increase the amount that class members could obtain under the BSWAT Payment Scheme [see Australian Government Solicitor magazine issue 2 for the case study].

  • Kamasaee v Commonwealth: In 2017, the Victorian Supreme Court approved a settlement involving a personal injury and false imprisonment damages claim, brought by nearly 2,000 persons transferred to Manus Regional Processing Centre in Papua New Guinea pursuant to the Migration Act 1958 (Cth) against the Commonwealth and its contract service providers.
  • AS v Commonwealth of Australia & Ors: In 2017, the Commonwealth resolved a personal injury claim on behalf of a minor, who was the representative plaintiff in a class action in the Victorian Supreme Court involving potentially 35,000 asylum seekers detained under the Migration Act at the Christmas Island Detention Centre between August 2011 and August 2014. The Commonwealth resolved the minor's claim, and therefore the entire proceedings, after successfully persuading the Supreme Court that the action should not continue as a class action.
  • Ibrahimi v Commonwealth: In 2017, the NSW Supreme Court dismissed a claim brought against the Commonwealth on behalf of the injured passengers and relatives of passengers who perished on board Suspected Illegal Entry Vessel (SIEV) 221 when it became shipwrecked off the coast of Christmas Island.
  • Giles v Commonwealth of Australia: In 2015, the NSW Supreme Court approved a settlement of a personal injury claim against the Commonwealth, the State of NSW and the Fairbridge Foundation, brought by 2 plaintiffs on behalf of a group who alleged that, as children, they were subjected to systemic physical and sexual abuse at the Fairbridge Farm School in NSW.

AGS is currently acting in a number of ongoing class actions, including:

  • Roo Roofing Pty Ltd v Commonwealth of Australia, which businesses have sued the Commonwealth in the Victorian Supreme Court over the termination of the Home Insulation Program
  • Brett Cattle Pty Ltd v Minister for Agriculture and the Commonwealth, in which the representative applicant has brought proceedings in the Federal Court on behalf of livestock producers, exporters and related industry participants, over the second suspension of the export of live cattle to Indonesia in June 2011
  • Smith v Commonwealth, in which the applicants have brought proceedings over the contamination of their land around the Williamtown airbase in NSW as a result of the use of aviation fire-fighting foam [see case study on PFAS contamination in this issue]
  • DBE 17 v Commonwealth, in which a minor has brought proceedings against the Commonwealth in the Federal Court, on behalf of persons who have been detained in accordance with the Migration Act between August 2011 and July 2017, alleging unlawful detention.

AGS has also seen the development of 'quasi class actions' where a claim is brought by many plaintiffs or multiple claims have been brought against the Commonwealth arising out of the same or similar factual situations, particularly in those jurisdictions where the State or Territory Supreme Courts have only limited rules and procedures to accommodate class actions. While not strictly 'class actions', these proceedings also pose similar practical and legal challenges when acting for the Commonwealth.

Challenges in class actions involving the Commonwealth

By their very nature, class actions have a number of unique features, including their size, particular court rules that apply and the process of ongoing supervision and case management by the Court throughout the duration of proceedings.

The class actions brought against the Commonwealth in recent times also raise issues of novel or new legal claims against the Commonwealth, for example alleging a duty of care that has not been previously tested or found to exist. Perhaps the most notable feature of class actions brought against the Commonwealth is their size and complexity. The claims often raise difficult issues of fact and law and require consultation and collaboration with multiple government agencies.

In some of the most recent proceedings in which AGS has acted, there have been thousands of members in the potential class. The size of the class can raise potential diversity of circumstances of different class members, though they must prove that they are related through common questions which are identified for determination.

The scale of representative actions, as compared with individual actions, affect many procedural steps, including:

1 Commencement of actions

An applicant or plaintiff commencing a class action must meet a number of threshold requirements, including:

  • a minimum of 7 persons
  • claims arising from similar or related circumstances, and
  • a substantial common issue of law or fact.

These threshold requirements mean that an applicant or plaintiff must invest considerable time into ensuring that the matter is suitable for a class action. It is essential if matters are to proceed as class actions that the definition of those who fall within the class is clearly identified and that there are common questions for determination which apply to all of the individual members. If there are not, there is the risk that the proceedings become inefficient or do not resolve the areas in dispute.3 An applicant or plaintiff (by their legal representatives) must take considerable care in establishing the substantial common issues of law or fact. Failure to do so may mean that the parties to an action that is not suited to being a class action are endlessly distracted with satellite litigation which unnecessarily delays consideration of the real issues in dispute. Further, in the absence of commonality, there is a risk that the Court will order that the proceedings should not continue as a representative proceeding, resulting in significant costs being thrown away.

An inevitable consequence for the Commonwealth of the threshold that must be crossed by an applicant is that once such an action is commenced, an applicant is much less likely to walk away from an action unless compensation has been paid.

2 Discovery and other interlocutory matters

The scale of class actions, and the involvement of at least 7 (and often many, many more) plaintiffs or applicants means that it is often more difficult to resist discovery processes in class actions for documents that do not relate to a specific plaintiff than it is in comparable single party actions.

In actions involving the Commonwealth, Courts will frequently be attracted to the concept of the discovery of documents describing government policies and practices which might inform the issues in dispute. A key priority for the Commonwealth is to restrict discovery to the agency (or agencies) that are directly involved in the issues in dispute. Even then, experience suggests that the sensitivity of discovered material will frequently involve other agencies, particularly when it is necessary to consult with agencies on restricting disclosure of information to avoid prejudice to the public interest in accordance with paragraph 7 of the Legal Services Directions 2017 (Cth).

In our recent experience, discovery in class actions has involved many hundreds of thousands of documents, with dozens of Commonwealth officers across many agencies involved in searching for, retrieving and assessing documents to be discovered.

Commonwealth lawyers must then assess the documents for relevance, privileges and sensitivity, with a further process of review by subject matter experts in relevant agencies to identify any prejudice to the public interest in the disclosure of documents. Despite the increasing scale of the discovery exercise, our experience is that very few documents are tendered or otherwise relied on in the trial of the ultimate proceeding. For class actions where the issues in dispute are complex, where there are possibly thousands of group members and where damage said to have been suffered by the group may date back decades, the number of documents that may be discoverable can be very significant. In these cases, it is easy for the process of discovery to become oppressive without careful management4 – both for the Commonwealth and also, for the applicant or plaintiff, who then has to review the material.

Courts will be reluctant to relieve the Commonwealth of discovery obligations only by reason of the volume of material alone. In many cases, large volumes of material are reviewed and ultimately discovered and produced by the Commonwealth. In these cases, AGS uses electronic document management technology to assist in the discovery process. These technologies can streamline the review process, create efficiencies and have other benefits in document management if the matter proceeds to trial. By way of example, in Kamasaee, AGS used document management technology to:

  • remove duplicates or near duplicates
  • eliminate or reduce the need to review individual documents by the use of visual analytic tools to identify clusters of irrelevant information
  • refine searches of repositories which were more likely to hold relevant information.

AGS relies on document management technology to assist with the preparation of documents for trial, including documents containing sensitive information. The parties retained a firm specialising in the conduct of electronic trials to assist with this process.

However, in some cases where the volume of material to be reviewed is oppressive or of limited relevance, orders for discovery can be resisted or, alternatively, limited in their reach. In Giles, the Commonwealth successfully resisted the plaintiff's application for discovery on the basis that the scope of documents sought to be discovered would have required the Commonwealth to review in excess of 20,000,000 documents and 3,382,470 files held by the National Archives and the Department of Immigration and Border Protection.5

3 Resolution of proceedings

Even the resolution of class actions is unique and complex. In most matters, the parties are able to reach settlement and enter consent orders disposing with the proceedings. This is not so for class actions for a number of reasons. As most class actions are 'significant matters',6 Commonwealth agencies must ensure approval by the Attorney-General or their delegate before the agency can enter into settlement negotiations. Even once settlement is agreed in principle, the rules of court require the approval by the court of any settlement (including if the settlement involves a discontinuance of the proceeding) to ensure that it is fair and reasonable and in the interests of the group members as a whole.7

Courts are very properly protective of the role they have in considering whether to approve the resolution of a class action. It will often be necessary for the parties, including the defendants, to make submissions, often on a confidential basis, on why the settlement is fair and reasonable for the group members as a whole.

The process of matters that run to trial for determination by the Court will also be more complex – in that event, the parties and the Court must have appropriate regard to the common questions that the Court will determine and precisely how the answers to those common questions will inform the assessment of related claims by class members in a way which maximises the utility of the initial trial. Failure to do so will lead to lengthy trials which determine so few common questions that each other group member's claim will face a further lengthy trial.

Pragmatic approach to class actions

Because of the unique nature of class actions and the possibility, if not managed properly, for them to become unwieldy and costly, AGS's experience has been that a pragmatic approach ensures the best outcome for all parties.

Early engagement with your legal team, including counsel, is essential. The priority is to ensure that the team appointed has sufficient experience in managing complex stakeholder relationships and is led by solicitors and counsel who have the ability to ensure a strategic approach which has appropriate regard to the many moving parts in such large-scale matters.

It's also essential that the core team meets on a regular basis to ensure those many moving parts are tracking appropriately at all times as the litigation progresses. Having that common team also allows for the team to be sufficiently flexible to adapt to the emerging needs of a litigation, conversely expanding and contracting as the matter develops at critical points, without losing the overall direction of the matter towards defence through a trial or resolution.

Consideration of interlocutory applications at the early stages of the proceedings, is also important. For example, strike-out applications can assist to refine the issues genuinely in dispute and to ensure unnecessary time and money is not spent on claims which are not tenable as a matter of unnecessary time and money is not spent on claims which are not tenable as a matter of law.8

Consideration should also be given at the earliest opportunity to whether the proceedings themselves are properly brought as a class action. This requires a consideration of the definition of the class members, whether there are common issues of fact and law, and also whether the proceedings will provide an efficient and effective means of dealing with the claims of the
class members.

Finally, consideration of security for costs application should be given where litigation funders are involved. Security for costs applications raise special issues in class actions. In the Federal Court and Victorian Supreme Court, there is an emphasis on the early and open disclosure of funding proposals. This is important for a number of reasons but also can inform whether an application should be brought, because Courts will be reluctant to impose security just because an applicant or plaintiff has insufficient funds.

In conclusion, the last few years has seen an increase in class actions against the Commonwealth, arising from a variety of circumstances, in both personal injuries and commercial disputes. While there are many challenges, there are advantages to the Commonwealth in class actions, including:

  • the costs of a class action will generally be less than litigating multiple related claims
  • the resolution of a class action is much more likely to achieve sustained finality of the result and avoid similar issues being re-litigated
  • greater coordination of Commonwealth resources in managing large-scale litigation and greater appreciation of risk and the challenges faced by the Commonwealth as a litigant
  • streamlining of claims arising from similar circumstances, with assistance of the Court's case management procedures.

1 Representative actions are also known as class actions or group proceedings. These terms are used interchangeably in this article.

2 The party bringing an action in a representative action in a State Supreme Court is the plaintiff, whereas that party is the applicant in proceedings in the Federal Court. In State Supreme Courts, the plaintiff sues the defendant, whereas the equivalent party in the Federal Court is the respondent. These terms are used interchangeably throughout this article, reflecting AGS's representative action practice experience in both State Supreme Courts and the Federal Court.

3AS v MIBP & Anor [2014] VSC 593.

4 See Giles v Commonwealth of Australia [2014] NSWSC 83 and AS v MIBP & Anor [2014] VSC 593.

5Giles v Commonwealth of Australia [2014] NSWSC 1531 at [19] per Garling J.

6 In accordance with the description as outlined in paragraph 3.1 of the Legal Services Directions 2017.

7 See s 33V of the Federal Court of Australia Act 1976, s 33V of the Supreme Court Act 1986 (Vic).

8AS v Minister for Immigration and Border Protection & Anor [2014] VSC 593; AS v Minister for Immigration and Border Protection & Anor [2016] VSC 351.


Back to contents page


Increasing your chances of successfully calling on an unconditional bank security

Unconditional bank guarantees, sometimes described as bank undertakings (both of which we refer to here as 'unconditional bank securities') are a form of security often used in construction contracts and commercial leases.

A hallmark of an unconditional bank security is the ability to convert it into cash on presentation to the relevant bank. While the terms of the security may be unconditional, often the contract or lease will qualify the right of the principal or lessor to call on the security. Recent authority has confirmed the unconditionally of the bank security itself; however, outlier decisions at the Supreme Court level remind us of the need to clearly detail in the contract or lease when the principal or lessor may call on the security.



Rachelle Hare
Simon Konecny

Rachelle Hare
Senior Lawyer
AGS Commercial
T 02 6253 7170


Simon Konecny
Deputy General Counsel
AGS Commercial
T 02 9581 7585


A promise to pay, with a contractual overlay

Unconditional bank securities are obtained by contractors or tenants from a bank and are commonly given to a principal or lessor as security for the due performance of the contract or lease. The security is an undertaking by the bank to the principal or lessor that the bank will pay it the amount of the security upon demand, often without consulting the contractor or tenant.

It is important to remember that the terms of a contract or lease may qualify the principal's or lessor's right to call on the unconditional bank security (at least so far as concerns the contractor or tenant). For example, the contract or lease may require the principal or lessor to give 3 days' notice in writing to the contractor or tenant of its intention to have recourse to the security. The contract or lease may also limit when the principal or lessor may call upon the security, even though the security itself may be unconditional.

Often, however, the contract or lease is silent on the question of when a principal or lessor can call on the unconditional bank security – this silence can itself lead to disputes between the parties.

When can a principal or lessor call upon an unconditional bank security?

The conventional rule is that a principal or lessor will not be restrained from calling on a bank security for an alleged breach where the claim is made in good faith.1 The rule is qualified by 3 exceptions, in that a principal or lessor cannot have recourse to the security if:

  • the recourse is fraudulent
  • the recourse is unconscionable
  • the principal or lessor has promised, either expressly or impliedly, under the contract that they will not call upon the security until it has been objectively determined that it is entitled to payment from the contractor or tenant (ie the contract or lease contains a negative stipulation).2

The principal's right to call on an unconditional bank security was firmly established in the Full Federal Court decision of Clough Engineering Ltd v Oil and Natural Gas Corporation Ltd (2008) 249 ALR 458, and it was most recently reaffirmed in the Supreme Court of Western Australia decision of CPB Contractors Pty Ltd v JKC Australia LNG Pty Ltd [2017] WASC 112. CPB Contractors confirmed that a beneficiary of an unconditional bank security can have recourse to it automatically and without having to establish that the existence of the breach is either agreed between the parties or beyond dispute. Provided that none of the above exceptions are triggered, a principal or lessor can approach the issuing bank and have the security paid out without question from the bank or without threat of an injunction from the contractor or tenant. Justice Le Miere, whose decision was upheld on appeal,3 emphasised that:

clear words will be required to support a construction which inhibits a beneficiary from calling on a performance guarantee where a breach is alleged in good faith, that is, non-fraudulently.4

Generally, the provisions of contracts or leases related to unconditional bank securities are considered to serve 2 purposes:

  • to provide security for a valid claim against the contractor or tenant
  • to allocate risk between the parties as to who shall be out of pocket pending the resolution of any dispute between them.5

Whether the unconditional bank security is intended to serve one, the other or both purposes is a question of construction of the relevant provisions of the contract or lease and the terms of the security itself.6 It can be critical to establish what the intended purpose of the security is, should a dispute arise between the parties as to when the principal or lessor is entitled to call upon the unconditional bank security.

A court is unlikely to find that a contract contains an implied negative stipulation if the purpose of the unconditional bank security would be defeated by restricting recourse to the security to only those instances where the breach is agreed or indisputable.7 While each case will depend upon subtleties of the particular contractual provisions, the following factors have been found to be consistent with a finding that the provisions of the contract or lease are intended to both provide security for the principal or lessor and serve the purpose of allocating risk of a dispute to the contractor or tenant:

  • where the provision requires the bank security to be an 'unconditional and irrevocable undertaking'8
  • where the provision requires the bank security to be 'payable on first demand'9
  • where the provision provides that recourse to the bank security can be at 'any time'10
  • where the unconditional bank security itself requires payment 'without proof of any breach or any other conditions and notwithstanding any contest or dispute'.11

While the recent decision of CPB Contractors supports the conventional view on how unconditional bank securities operate, there are other cases which illustrate the difficulties that may arise in the absence of an express provision in a contract relating to what the purpose of the bank security is, or to the circumstances in which recourse might be had to the bank security. Walton Construction Pty Ltd v Pines Living Pty Ltd [2013] ACTSC 237 and Lucas Stewart Pty Ltd v Hemmes Hermitage Pty Ltd [2010] NSWCA 283 are both construction contract cases in which the court found that a principal's entitlement to have recourse to a guarantee was conditional upon the objective fact of the contractor's non-compliance. In those cases, it was not enough that the principal asserted it had a claim for breach; that breach had to be agreed or objectively determined before recourse could be had to the guarantee.

A similar conclusion was reached in Universal Publishers Pty Ltd v Australian Executor Trustees Ltd [2013] NSWSC 2021, a commercial lease matter, where the court held that if a tenant could demonstrate that there was a serious question to be tried as to the existence of the breach, then the landlord would be restrained from accessing the guarantee. In each of those cases, it was significant to the courts' reasoning that, upon construing the contract, it appeared the bank security was provided solely as security for performance of the contractor's or tenant's obligations (rather than for the purpose of allocating risk as between the parties). This contradicts the approach established in Clough and reaffirmed in CPB Contractors.

The inconsistency demonstrated by the different approaches adopted by the courts highlights how important it is for parties to make sure that their intentions are clearly reflected from the outset in their contract or lease provisions dealing with unconditional bank securities.

How to draft your contractual security provisions

Consider amending your contract or lease provisions in the following manner:

  • Expressly address the circumstances in which recourse may be had to the unconditional bank security.
  • The unconditional bank security should be stated to be 'unconditional and irrevocable'.
  • Clarify when recourse may be had to the unconditional bank security, as follows:
    • The principal or lessor may call upon the security 'at any time and without the requirement to provide prior notice' to the contractor or tenant.
    • The security will be 'payable on first demand'.
    • Recourse may be had by the principal or lessor to the security 'in the event of an alleged breach'.

It is preferable for the parties to have the discussion about unconditional bank securities prior to execution of a contract or a lease, and to agree the approach to take at that time, rather than face a lengthy court battle over injunctions and interpretation of the relevant contract or lease provisions at a later date.


1Clough Engineering Ltd v Oil & Natural Gas Corp Ltd (2008) 249 ALR 458 ('Clough').

2CPB Contractors Pty Ltd v JKC Australian LNG Pty Ltd [2017] WASC 112 at [68] ('CPB Contractors').

3CPB Contractors Pty Ltd v JKC Australia LNG Pty Ltd [2017] WASCA 123.

4 Ibid at [70] citing Clough at [83].

5Clough [79].

6CPB Contractors at [70]; Cf Lucas Stewart Pty Ltd v Hemmes Hermitage Pty Ltd [2010] NSWCA 283 at [43].

7CPB Contractors at [82].

8CPB Contractors at [76]; Fletcher Constructions Australia Ltd v Varnsdorf Pty Ltd [1998] 3 VR 812.

9Bachmann Pty Ltd v BHP Power New Zealand Ltd [1999] 1 VR 420.

10CPB Contractors at [77].

11CPB Contractors at [76]; Clough at [100].


Back to contents page


Technology-assisted review

Discovery is often the most costly stage of litigation.1 The mere mention of discovery invokes images of junior lawyers locked away in dimly lit rooms for weeks or months reviewing dusty boxes of documents. Modern discovery is a different species, however, regularly involving the review of millions of electronic documents – a result of the ease by which these records are now generated and stored. The proliferation of email has meant that informal messages that were once relayed verbally now take written form, and the low cost of electronic storage means that these records persist, further complicating discovery.

The rise in electronic communication, together with a burgeoning class action sector and growth in complex litigation, presents a challenge for both practitioners and courts. This environment threatens one of the courts' overriding objectives, being the speedy and efficient administration of justice and resolution of disputes.2

Against this background, technology-assisted review (TAR) is a useful tool which, if embraced by the profession, has the potential to significantly reduce the time and cost associated with manual document review, and which may also limit the inconsistencies caused by human error.



Claudia Oakeshott
Christopher Henies
Lara Strelnikow

Claudia Oakeshott
Senior Lawyer
AGS Dispute Resolution
T 02 6253 7285

Christopher Henies
AGS Dispute Resolution
T 02 6253 7349

Lara Strelnikow
Graduate Lawyer
AGS Dispute Resolution
T 02 6253 7461



The decision of Vickery J in McConnell Dowell Constructions (Aust) Pty Ltd v Santam Ltd & Ors3 (McConnell Dowell No. 1) illustrates how the volume of discovery in large-scale litigation renders traditional review processes inefficient, costly and unwieldy when compared with TAR.

With the use of TAR in that case, the number of documents identified by the plaintiff as relevant had already been reduced from approximately 4 million to 1,400,000.4 In approving the use TAR, Vickery J relied upon decisions from other jurisdictions including the United Kingdom, Ireland and the United States.5 Referring to the English case Pyrrho Investments Limited v MWB Property Limited,6 Vickery J observed that the use of de-duplication had reduced the 17.6 million documents involved down to approximately 3.1 million.7 His Honour found that the further use of TAR would dramatically reduce the number of documents to be reviewed and, conversely, that a traditional review would be neither cost-effective nor proportionate:

'… employing a traditional manual discovery process can work to place the cost-benefit of conducting litigation in a large document case at serious risk.'8

What is TAR?

Although sometimes considered synonymous with predictive coding, TAR is an umbrella term encompassing all methods in which technology can be deployed to assist with document review (including predictive coding). For example:

  • de-duplication: the identification and subsequent removal of documents with identical content
  • email threading: the grouping of emails from the same conversation so that they can be reviewed as one document
  • clustering: a method where documents with similar words are grouped together
  • near de-duplication: the grouping of similar, almost identical, documents (similar to clustering)
  • concept searching: a search method that retrieves results based on the ideas expressed in a document as opposed to traditional Boolean keyword searches.

Predictive coding

Predictive coding uses algorithms that can be trained to analyse and code documents for likely relevance or privilege. It does not obviate the need for human review, which remains important, particularly early on when the algorithm is being trained.


Two useful measures of the efficacy of predictive coding are 'recall' and 'precision':

  • 'recall' is a measure of how well the algorithm retrieves relevant documents (for example,
    a 75% recall rate means that 75% of relevant documents are retrieved and 25% are missed)
  • 'precision' is a measure of how many irrelevant documents the algorithm retrieves (a 75% precision rate suggests that 25% of documents retrieved are not truly relevant).

'Recall' is said to be a measure of completeness, where 'precision' is a measure of accuracy. Generally, both cannot be achieved at the same time: one must be traded off against the other.9

Predictive coding protocols are usually described as falling within 3 categories: simple passive learning (SPL), simple active learning (SAL), and continuous active learning (CAL), in increasing order of complexity.

  1. Simple passive learning: SPL begins with a randomly selected seed set of documents which are reviewed and coded by a human reviewer as relevant or not relevant (or privileged or not privileged). This coding is then used to train the algorithm to build a scoring system which ranks other documents from the broader population along a spectrum of relevance or privilege (for example, on a scale of –1 to 1, where documents with a value of close to –1 are least likely to be relevant, documents close to 1 most likely to be relevant, and documents clustering around 0 to be most marginal). The documents ranked most likely to be relevant will be evaluated to see how 'stable' the algorithm is. If the results are inadequate, more documents will be introduced to further train the system until the algorithm is sufficiently 'stable'. Human reviewers then review and code those documents ranked by the algorithm above a certain threshold, depending on what recall rate is sought.10
  2. Simple active learning: The SAL protocol differs in that the initial seed set does not begin with a random selection of documents; rather, it begins with documents identified as likely to be relevant (usually by keyword search). Another divergence is that as the algorithm is trained, it pulls documents to the seed set which it is most unsure about rather than randomly (so in the above example, documents closest to a value of 0) for manual review and coding. Once the algorithm is sufficiently robust, the process then proceeds in the same way as SPL.

Both the SPL and SAL protocols are initially iterative but become finite when documents are either selected for human review or excluded. By contrast, CAL (as the name implies) remains iterative.

  1. Contiuous active learning: The CAL protocol begins, again, with an initial seed set which is coded by a human reviewer and used to generate a ranking system for the whole population of documents. A further sample of the top ranked documents from the whole population is then coded by a human reviewer but, unlike SAL and SPL, once coded these documents are then fed back into the algorithm to re-train and improve it. The process is repeated such that human judgments are continuously fed into the system and the whole population of documents is repeatedly re-ranked.

The CAL protocol has been demonstrated to require less human review to achieve a specified recall rate when compared with SAL or SPL.11 Consequently, it is now the generally preferred method for predictive coding.

In all cases, ideally, the initial seed sets should be reviewed by a senior lawyer 'who has mastered the issues in the case',12 to ensure accuracy and consistency.

Judicial acceptance

Jurisdictions outside of Australia have embraced both TAR and predictive coding. In the US case of Rio Tinto Plc v Vale SA, Magistrate Judge Peck stated that 'it is now black letter law that where the producing party wants to utilize [TAR] for document review, courts will permit it.'13

Within Australia, the Supreme Court of Victoria's Practice Note specifically provides that TAR 'will ordinarily be an accepted method of conducting a reasonable search … when there are a large number of Electronic Documents to be searched and the costs of manually searching the documents may not be reasonable and proportionate.'14 Even before the Practice Note came into effect, on 2 December 2016, Vickery J delivered the first Australian judgment approving the use of predictive coding for the purpose of carrying out discovery in McConnell Dowell No. 1.

While there is no equivalent Commonwealth practice note (yet), the use of TAR is consistent with the principles which guide the courts in making orders for discovery, being the just and efficient resolution of disputes and ensuring that documents sought and produced in discovery are significantly probative in nature. Recent cases confirm that courts are likely to be receptive to the use of TAR.15

How can TAR be used in proceedings?

There are case examples where all parties involved in proceedings have agreed to and collaborated on the use of TAR.16 In such cases, parties may jointly approve a 'predictive coding protocol' and agree, for example, on the appointment of a joint operator,17 on the criteria for inclusion of documents, and the target 'recall' rate. Where predictive coding is approached collaboratively, all the parties may even review documents to determine their relevance and thereby contribute to training the algorithm.18

Unsurprisingly, there are also decisions where parties have disagreed about using predictive coding and/or TAR, or the proposed method for its use.19 Courts may also order discovery by TAR whether or not the parties consent to it.20

As observed by Vickery J in McConnell Dowell No. 1, TAR does not have to be used by all parties involved in the proceedings: a single party can use the technology for its own review.21

Issues in TAR

Efficient use of resources

One of the primary benefits of TAR and predictive coding is cost-effectiveness. As noted above, discovery is one of the most costly aspects of litigation. However, there are some time-consuming aspects of predictive coding, particularly in the development stage (for example, agreeing a protocol and training the algorithm). An assessment should be made in each case whether there is a sufficient volume of documents, and hence that there will be later time-saving, to justify the establishment costs. In addition, there needs to be a large enough seed set (usually, 500–1,000 documents) for the predictive coding algorithm to learn from and to apply to a broader population. A larger seed set inversely correlates with margin of error.

In AGS's experience, as a general proposition, at least 5,000–10,000 electronic documents are required for predictive coding to be worthwhile. However, the Supreme Court of Victoria's Practice Note encourages the more general use of technology for matters where the discovery exercise is likely to be significant, specified as involving volumes of only 500 documents or more.22 While advanced predictive coding tools may not be cost-effective in these smaller matters, parties can still benefit from the use of other TAR tools, such as email threading and

Inadvertent disclosure

Although each of the SPL, SAL and CAL predictive coding protocols contemplates that all documents produced are reviewed by human reviewers,23 if there is to be minimal manual review undertaken following the application of TAR, there is a risk of inadvertent disclosure of privileged or confidential material.24 A protocol agreed between the parties can mitigate against this risk (for example, setting out terms for the return and destruction of documents).

Enhanced quality of the discovery processes

Many of the objections raised in relation to the use of this emerging technology question the reliability and accuracy of the review process; yet, traditional human review is far from perfect. As Fullam J noted in the Irish case of Irish Bank Resolution Corporation Ltd & Ors v Quinn & Ors,25 predictive coding is at least as accurate, if not more accurate, than manual review processes in discovery involving numerous documents:

[Even if ] one were to assume that TAR will only be equally as effective, but no more effective, than a manual review, the fact remains that using TAR will still allow for a more expeditious and economical discovery process.26

Unlike human reviewers, a computer is consistent, does not tire or grow bored of reviewing documents, does not need rest breaks, and can far more rapidly process records.

In any event, the predictive coding protocols described above do not remove lawyers from the review process – lawyers engage with the technology throughout.

Model litigant obligations

A further compelling rationale specific to Commonwealth agencies for using TAR in discovery relates to our responsibilities as model litigants; specifically, our obligation to act honestly and fairly in handling claims and litigation, including by keeping the costs of litigation to a minimum.27 Arguably, using a technology like predictive coding, which can minimise delay, reduce costs and enhance the quality of discovery, is consistent with this obligation.


The use of TAR will not always be appropriate – for example, where the document pool includes large numbers of hard copy or handwritten documents, drawings or data sets.28 However, in the context of litigation involving extensive electronic records, it will become increasingly difficult to justify a decision to rely on traditional manual review given the resulting increased time, expense and the likelihood of inconsistent results.

AGS has been using TAR to assist in document management in a number of our matters. Most recently, the ACCC (with AGS acting) filed applications in the matters of ACCC v Volkswagen Aktiengesellschaft & Anor (NSD1462/2016) and ACCC v Audi Aktiengesellschaft & Anor (NSD322/2017) seeking orders for the appointment of a referee to inquire into whether predictive coding should be used to streamline future discovery processes in those proceedings, and, if so, what protocol should be adopted.


  1. Federal Court of Australia, Technology and the Court Practice Note (GPN-TECH), [3.2]; see also the comments of the Australian Law Reform Commission, 'Managing justice: A review of the federal civil justice system' (Report 89, 2000) [6.67].
  2. Section 37M of the Federal Court of Australia Act 1976 provides, among other things, that the overarching purpose of the civil practice and procedure provisions (ie the Federal Court Rules 2011 and Federal Court and Federal Circuit Court Regulation 2012) is to facilitate the just resolution of disputes…as quickly, inexpensively and efficiently as possible. Section 37N requires that the parties must conduct the proceeding in a way that is consistent with the overarching purpose.
  3. [2016] VSC 734.
  4. McConnell Dowell Constructions (Aust) Pty Ltd v Santam Ltd & Ors [2016] VSC 734, [3] (Vickery J) (McConnell Dowell No.1).
  5. Ibid [18]–[25] (Vickery J); see also, Pyrrho Investments Limited v MWB Property Limited [2016] EWHC 256 (Ch); Irish Bank Resolution Corporation Ltd & Ors v Quinn & Ors [2015] IEHC 175; Rio Tinto v Vale 14 Civ 3042 (RMB)(AJP) (2 March 2015).
  6. [2016] EWHC 256 (Ch).
  7. McConnell Dowell No. 1, [19] (Vickery J).
  8. Ibid [6].
  9. See McConnell Dowell Constructors (Aust) Pty Ltd v Santam [2017] VSC 640 (McConnell Dowell No. 2), [21]–[22] (Vickery J).
  10. Deciding a recall rate involves balancing the desirability of completeness against the burden of work, where a recall rate only
    5% higher, for example, may correlate with substantially more effort, time and expense.
  11. Grossman M. and Cormack G., 'Comments on "The implications of rule 26(g) on the use of technology-assisted review" ', 7 Federal Courts Law Review 286, 297 (2014).
  12. Pyrrho Investments Limited v MWB Property Limited [2016] EWHC 256 (Ch) at [20].
  13. 1:14-cv-3042 (SDNY March 2, 2015).
  14. Supreme Court of Victoria Practice Note, [8.7].
  15. There has not yet been any case at the Commonwealth level expressly approving the use of TAR; but, for example, in Money Max Int Pty as Trustee for the Goldie Superannuation Fund v QBE Insurance Group Limited Murphy J ordered the Respondent, who had used TAR for the purpose of giving discovery, to provide a report detailing the methods it had applied.
  16. See for example, McConnell Dowell No.1 and No. 2.
  17. The Supreme Court of Victoria Practice Note defines joint operator at paragraph [3.1] as 'a person, organisation or firm experienced in the use of technology assisted review and with access to the necessary software to manage electronic discovery for the parties'.
  18. See the case management order 'Protocol relating to the production of Electronically Stored Information ("ESI") issued in In Re: Actos (Pioqlitazone) Products Liability Litigation', No. 6:11-md-2299 (WDLa July 27 2012), and McConnell Dowell No. 2.
  19. See US case law, for example: Global Aerospace Inc. v Landow Aviation, LP, No. 61040 (Loudoun County, Va. Ct. Apr. 23, 2012); Monique da Silva Moore, et al v Publicis Groupe SA & MSL Group, No. 11 Civ. 1279 (ALC) (AJP) (SDNY Feb. 24, 2012).
  20. Supreme Court of Victoria Practice Note, [8.7].
  21. McConnell Dowell No. 1, [31].
  22. Supreme Court of Victoria Practice Note, [8.3].
  23. The risk with each of these protocols is instead that relevant documents are not reviewed and not discovered (which can but minimised to some extent with QA processes and setting a higher recall rate).
  24. See for example, the High Court decision relating to inadvertent disclosure involving electronic methods of discovery and legal professional privilege in Expense Reduction Analysts Group v Armstrong Strategic Management and Marketing (2013) 250 CLR 303. This decision did not involve predictive coding.
  25. [2015] IEHC 175.
  26. Ibid, [66]–[67].
  27. Legal Services Directions 2017 (Cth), app B, cl 2(e).
  28. Erick Gunawan and Tom Pritchards, 'Technology assisted review', Technology and the Law (June 2017) <…;.


Back to contents page

Constitution s 44

Disqualified senators

In early 2017 the High Court, sitting as the Court of Disputed Returns, held 2 persons elected as senators in 2016 to be disqualified from being chosen or sitting as senators. The first was Rodney Culleton, disqualified pursuant to s 44(ii) of the Constitution on the basis that at the time of his nomination and election he had been convicted, and was subject to be sentenced, for an offence punishable by imprisonment for at least 1 year. The second was Robert Day, disqualified pursuant to s 44(v) of the Constitution on the basis that he had a pecuniary interest in an agreement with the Public Service of the Commonwealth.

AGS acted for the Attorney-General, who was joined as a party in each of the proceedings.


Andrew Chapman
Nerissa Schwarz

Andrew Chapman
Senior General Counsel
Office of General Counsel
T 02 6253 7206


Nerissa Schwarz
Senior Lawyer
Office of General Counsel
T 02 6253 7078


Re Culleton (No 2) [2017] HCA 4

The Culleton case concerned a discrete legal question: whether a conviction, subsequently annulled, operated to disqualify a person under s 44(ii) of the Constitution. Section 44(ii) disqualifies from being chosen, or sitting, any person who '… has been convicted and is under sentence, or subject to be sentenced, for any offence punishable under the law of the Commonwealth or of a State by imprisonment of one year or longer'.

The joint judgment of Kiefel CJ, Bell, Gageler and Keane JJ held that, properly construed, the Crimes (Appeal and Review) Act 2001 (NSW), pursuant to which Mr Culleton's conviction was annulled, operated with prospective effect only. It did not retrospectively annul his conviction and remove his disqualification under s 44(ii). Their Honours did not consider whether a retrospective annulment of a conviction could remove a disqualification. Only Nettle J dealt with this issue, holding that retrospective annulment would not remove a s 44(ii) disqualification: the need for certainty in the electoral process required that s 44(ii) be engaged by a conviction in fact, even if later annulled.

Re Day (No 2) [2017] HCA 14

The Day case concerned complex factual and legal questions. Mr Day wished to establish his electorate office at premises at 77 Fullarton Road, Kent Town, South Australia owned by a corporation which was the trustee of a discretionary trust known as 'the Day Family Trust'. Mr Day was one of the beneficiaries of that trust.

From late 2013, Mr Day believed that the Commonwealth was unwilling to take a lease of part of the Fullarton Road property for use as his electorate office so long as an entity in which he had an interest owned the freehold. Accordingly, in April 2014 B&B Day sold the Fullarton Road property to another company, Fullarton Investments Pty Ltd. Fullarton Investments was also a trustee company, being the trustee of a discretionary trust known as 'the Fullarton Road Trust'. The sole director of this trustee company (Mrs Smith) was the wife of a business associate of Mr Day (Mr Smith). B&B Day was one of the beneficiaries of the Fullarton Road Trust.

The nominal purchase price for the sale of the Fullarton Road property was $2.1 million. No money actually changed hands. Instead, B&B Day and Fullarton Investments executed a document that acknowledged the provision of a 'vendor finance' loan by B&B Day to Fullarton Investments in the sum of $2.1 million. Under an arrangement between Mr Day, Mr Smith and Mrs Smith, Fullarton Investments 'would simply hold the Fullarton Road property and collect rent on a regular basis'. That rent would then 'pass back to the Day Family Trust'.

Eventually, on 1 December 2015, a lease of the Fullarton Road property was entered into between Fullarton Investments and the Commonwealth. Pursuant to the lease, on 26 February 2016 Fullarton Investments nominated a bank account in the name of 'Fullarton Nominees' for receipt of rent. Fullarton Nominees was a business name owned by Mr Day and the bank account was his. As it turned out, the Commonwealth did not in fact pay any rent under the lease.

At all relevant times there was a loan facility provided by the National Australia Bank (NAB) to B&B Day and other companies with which Mr Day was associated. This loan facility was secured by a mortgage over the Fullarton Road property (even after the sale to Fullarton Investments) and by a guarantee and indemnity given by Mr Day and his wife.

The Court unanimously held that Mr Day was ineligible to have been elected at the 2016 election by reason of s 44(v) of the Constitution, which disqualifies any person who has '…any direct or indirect pecuniary interest in any agreement with the Public Service of the Commonwealth otherwise than as a member and in common with the other members of an incorporated company consisting of more than twenty-five persons'. All members of the Court held that the fact that Mr Day was the owner of the bank account nominated as the recipient of the rental monies for the lease with the Commonwealth was sufficient to engage s 44(v) (with Gageler J and Nettle and Gordon JJ holding that a disqualifying interest in the lease also arose because Mr Day was exposed in other ways to the possibility of a financial gain or loss from the performance or breach of the lease).

All 7 Justices, in 4 separate sets of reasons, held that s 44(v) has a wide purpose and operation, declining to follow an earlier decision of former Chief Justice Barwick in Re Webster (1975) 132 CLR 270 in which his Honour held that s 44(v) had the limited purpose of preventing the influence of parliamentarians by the executive. All 7 Justices expressed the view that part of the purpose of s 44(v) is to protect against the personal interest of a parliamentarian influencing the exercise of their public duties. In 3 separate sets of reasons, Kiefel CJ, Bell and Edelman JJ and Gageler J and Keane J also suggested, with some variation, that despite the provision's breadth, it would not cover certain routine agreements for benefits provided by the Commonwealth that were available to the community generally.

Filling the vacancies

In each case the Court ordered a special count of the Senate ballot papers as if Mr Culleton and Mr Day were removed and their preferences distributed to the next preferenced candidates accordingly.