Legal information is the lifeblood of our system of justice, and the broader liberal-democratic polity in which our legal institutions form a critical part. It is not enough that this information exists somewhere: it must be readily available, intelligible, accessible to those who want it, and it must be reliable.

While AustLII is the embodiment of positive progress to achieve this goal, we face new threats that could result in new barriers to access the information needed for the effective operation of our justice system. Put simply, without free access to legal information, the rule of law, and the human rights that rely on it, are themselves in jeopardy. Lawyers' most fundamental duty is to safeguard the rule of law, and so the way we respond to these threats will go a long way in shaping the legal profession for the remainder of the 21st century.

The most pressing threat to the free access of legal information in 2025 relates to its reliability. The challenge to ensure the reliability of legal information is both greater and more complex than when AustLII was formed. Previous generations focused on ensuring that legal information - legal rules, judgments, and authoritative analysis - would not remain locked away behind doors whose key was available only to those with significant financial resources and specialised knowledge. Now, thanks to the movement in which AustLII has been a leading light, many of those doors are open.

The rise of artificial intelligence, however, threatens this progress. As increasing proportions of the unimaginably vast forest of human-created information became available to anyone with an internet connection, the imperative shifted towards the need to navigate through this forest. The advent of powerful search engines in the early 2000s made a significant contribution to address this issue. But only with the rise of generative AI, with its targeted responses to our specific questions, did a magic-seeming forest guide appear.

Unfortunately, of course, this progress has coincided with a rise of 'false facts' - AI hallucination, mis- and dis-information. In this presentation, I will lay out some critical aspects of this challenge and offer some suggestions of how we in the legal profession should respond.

Artificial Intelligence presents many opportunities and challenges.

The Federal Court of Australia has concluded its open consultation with the legal profession about the use of AI technologies in court proceedings and is in the process of developing official guidance.

His Honour's presentation will explore the important professional and ethical responsibilities of legal practitioners and litigants in person whilst balancing the risk of AI against the advantages of its use, including those which relate to access to justice.

The digital search landscape is undergoing a profound transformation with the integration of Artificial Intelligence (AI) and Large Language Models (LLMs) into online search platforms like Google, leading to the prominence of AI Overviews and advanced Featured Snippets. For Legal Information Institutes (LIIs) worldwide, this evolution presents a critical decision: to embrace this emerging reality to assure that AI Overviews and Featured Snippets results are accurate and authoritative for the public, or to prevent companies like Google from effectively including our content into their results to prevent them from benefiting from our previous expenditure of significant resources to create that content and put it online. The Legal Information Institute at Cornell Law School has chosen the first approach.

The purpose of this document is to outline the strategic recalibration we have embarked upon to preserve and enhance the public accessibility of our resources in an increasingly AI-mediated information environment. Since AI-driven search shifts user behavior towards "no-click searches" where Google itself answers user queries, we are working to move beyond solely driving clicks to our law.cornell.edu website.

AI's increasing sophistication in understanding human language and intent has reshaped Search Engine Results Pages, requiring us to adapt to remain visible and impactful. To see our content populate Google's AI Overviews and Featured Snippets, we have needed to learn about Answer Engine Optimization (AEO) while continuing to pay attention to traditional Search Engine Optimization (SEO), which remains fundamental.

The presentation and accompanying paper begin with a detailed explanation of Google AI Overviews and Featured Snippets, highlighting their differences and the strategic shift towards AI dominance and the "no-click search" trend. The work next explains our view that AEO is crucial for LIIs to maintain visibility and authority, provide direct answers, establish brand authority, and adapt to changing user behavior in an AI-dominated search environment. It then describes strategies for optimizing AEO, including newer concepts such as prioritizing E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) and adopting user-intent focused content, while emphasize the continuing need for mastering structured data and technical SEO fundamentals, enhancing user experience and site performance, and evolving measurement frameworks beyond traditional metrics. It concludes by embracing a dual optimization approach of leveraging strong traditional SEO fundamentals with targeted AEO strategies to secure our continued relevance and impact in the dynamic future of search, ensuring legal information remains both accurate and freely and easily discoverable to the general public.

The principle of Free Access to Law has guided the work of Legal Information Institutes for over 30 years. Its early success was made possible by the availability of low-cost electronic publishing via the World Wide Web. This paper discusses the challenges raised by emerging GenAI and LLM technologies. It examines the available responses and makes suggestions for the best way forward.

The European Union is built on the coexistence of multiple legal systems and languages, yet there is no comprehensive, unified legal information infrastructure that reflects this diversity. Existing legal databases are fragmented by jurisdiction and typically monolingual, often covering only national law or EU law in isolation. This fragmentation limits accessibility, hinders transparency, and creates barriers for both legal professionals and the development of AI-based legal applications that depend on structured, interoperable legal data across borders.

The intended paper shall present a concept and working prototype for a multijurisdictional, multilingual legal knowledge graph that integrates EU law, national law from all member states, and relevant court decisions. The knowledge graph is designed to be language-independent, while preserving legal precision through structured representations of norms, relationships, and authorities. It employs a semantic framework that supports legal cross-referencing, multilingual concept linking, and the accurate alignment of parallel legal structures across jurisdictions.

Unlike traditional legal databases, the graph is open and interoperable, enabling the reuse of legal knowledge as structured data. This makes it an ideal foundation for building a new generation of AI tools, such as legal search assistants, compliance checkers, decision support systems, and workflow automation agents. These applications can interact with the graph using natural language, rule-based logic, or hybrid AI techniques, offering both legal professionals and public institutions scalable access to trustworthy and contextualized legal information.

The system also enables the linking of private legal documents (e.g., contracts, administrative rulings) to the public legal graph in a secure and logically coherent way. This allows legal practitioners to build powerful tools that align private legal content with applicable norms and precedents. Importantly, the graph can be extended with metadata, annotations, and ontologies, facilitating collaborative enrichment and domain-specific adaptations.

In summary, the intended paper will argue for the necessity and feasibility of an open, pan-European legal knowledge graph as the foundational infrastructure for Legal Informatics in the EU. By providing a multilingual, multijurisdictional representation of the law in structured, machine-readable form, the proposed system opens the path to meaningful digital transformation in the legal domain, supporting both access to justice and the responsible use of AI in legal practice and governance.

Kotahitanga represents the concept of togetherness, unity and solidarity - working together to lift each other up. Kotahitanga recognises diversity of parties, that singularly one can achieve, but working together creates magic.

In 2005 Donna Buckingham called NZLII an attempt at a magic bullet – a single web site to access New Zealand legal information for free. Since then, NZLII has served up needed legal information to millions of users. The original plan was for NZLII to be hosted on AustLII (with a NZLII front door) until a critical mass was reached in terms of coverage, triggering the ultimate goal of achieving data sovereignty and hosting it within Aotearoa New Zealand.

Twenty years later, we are still with AustLII, having realised that together we are stronger. Kotahi te aho ka whati, ki te kāpuia e kore e whati – “one strand of flax is easy to break, but many strands together will stand strong" – i.e. unity and collaboration result in greater strength and resilience than individual effort.

In light of those principles, the opportunity to do a major revamp meant that we could take a hard look at how NZLII can continue to serve our community of users. This talk will discuss and demonstrate the key aspects of the redesign; explain why they are important to our user community and how they will contribute to the sustainability of NZLII. It will also look at how we can continue making changes to stay relevant in an environment where the focus is on AI and LLM.

In this exciting time of redevelopment and reimaging of NZLII, we are very grateful for the ongoing manaakitanga shown to it by AustLII and all their amazing staff.

Ko te puāwaitanga o ngā moemoeā, me whakamahi

The Pacific Islands Legal Information Institute (PacLII) aims to provide a comprehensive portal of legal databases for 20 Pacific Island jurisdictions, including current versions of laws that incorporate all amendments. However, the consolidation of these laws and ensuring access to those collections has been challenging across the region.

This presentation will examine different approaches jurisdictions use to consolidate their statute law and how these collections are made accessible. Releasing up-to-date laws for republication serves the public interest as well as supports cross-jurisdictional legal research and the development of jurisprudence. A key consideration in maintaining legislative currency is long-term sustainability. By leveraging an economy-of-scale approach, PacLII, as an established and experienced legal publisher in the Pacific, is well positioned to deliver a low-cost legislative consolidation service that enhances the currency, reliability, and point-in-time accessibility of Pacific laws, which can then be shared as open data.

Cylaw was founded in January 2002 as a personal initiative and became the main source of free legal information in Cyprus. In 2011, the Cyprus Bar Association took over the operation of Cylaw while upgrading services and expanding database so that it includes all the primary sources of Cypriot law. This presentation will briefly introduce Cylaw’s origins, explain its guiding philosophy, and outline future plans.

As AI transforms how people search for and interact with the law, the question of government infrastructure for support of open law services becomes pressing. This talk focuses on open legislative data services and the growing demand for bulk, machine-readable legislation to power public-facing AI tools.

The presentation will explore emerging use cases where AI depends on free legislative data, from chatbots that answer everyday legal queries to tools that support policy analysis and compliance. The discussion highlights the evolving role of drafting and publishing offices: not just as legislative publishers, but as stewards of reliable data pipelines and interfaces.

Topics explored include accuracy, trust, privacy, timeliness, and reuse in information supply chains, and the shift from document-centric to fragment-based models better suited to AI systems.

This presentation will outline recent work by Jersey's Computer-Readable Legislation Project on capturing and visualising the logical structure of legislation. We have particularly been working with a team from Singapore on their "L4" language and a logical pathways visualiser. We have been looking for ways in which legislative drafters can, without tech ability, represent the if-and-or-not structure (propositional logic) of their drafts in a version of the text that can be automatically converted into L4. This L4 translation therefore only captures the if-and-or-not structure (reflecting just propositional logic, rather than predicate logic). So it does not use the full capabilities of L4, but it can do useful things in the visualiser that the Singapore team developed.

Ideally, we want the drafter's if-an-or-not version to be also automatically convertible into other computer languages such as AustLII's yScript (for their DataLex consultation system, via their yLegis preprocessor), and possibly Logical English and Blawx among others. The idea is to show legislative drafters what can be done if the logical structure is captured, as a useful tool so far as it goes but also as a taster to get them to think about whether they want to get further technical help for languages like L4 and Blawx or whether they want to change their drafting style to be able to use DataLex or Logical English directly.

The point of all of these approaches is to help users to understand how legislation applies to their scenarios, by guiding them through the often complex logical structures on which it is built. But the point is to do so in a way that respects the wording of the legislation and does not stray into interpreting substantive expressions in the text (leaving the human user to work out what they mean instead), and is also practical for legislative drafters to implement.

The digitization of legislation has advanced worldwide, yet national and international standards remain divergent. In Japan, the Japan Legal Information Institute (JaLII) designed the Japanese Legal Standard XML Schema (JLS), later adopted in the government's e-LAWS system. Internationally, Akoma Ntoso (AKN) has become a widely used schema. Both JLS and AKN advanced legal informatics, but their incompatibility has limited interoperability and comparative law research.

This paper shows how JLS has been extended through two projects toward an international infrastructure for comparative law. The first project, Development of Major Applications for Making Effective Use of the Japanese Statutes as Open Data (2019–2022), established rewriting rules to convert JLS into AKN and adapted open-source tools such as the Lime editor and XML converters. This demonstrated the feasibility of schema conversion and laid a foundation for secondary use of Japanese statutes.

The second project, Construction of a Comprehensive Database of Corresponding Provisions for Comparative Legal Studies (2024–2028), develops methods for identifying provision-level correspondences across jurisdictions. A recent study [3] proposed a one-to-many mapping approach, tested on family law provisions in the Japanese and German Civil Codes. While accuracy remains modest (best F1-score = 0.3476), the approach improved coverage beyond earlier methods.

By integrating schema conversion with provision mapping, this research demonstrates how a domestic XML schema can contribute to a multilingual infrastructure for comparative law.

The Netherlands has developed a unique public service that delivers personalized digital notifications about regulations, announcements, and administrative procedures relevant to a citizen's direct living environment. This service marks the culmination of a 30-year transformation from paper-based publication to fully digital, legally authentic government information.

The country's four layers of government—national, provincial, municipal, and water authorities—each publish in their own official journal, all using a unified national infrastructure managed by the central Publications Office. Since the 2021 Promulgation Act (Bekendmakingswet), all regulations and announcements are digitally published in journals such as the Staatsblad, Staatscourant, Provinciaal blad, Gemeenteblad, and Waterschapsblad.

Previously, announcements appeared in regional newspapers, which was costly and often overlooked. Article 20 of the Promulgation Act now authorizes the Minister of the Interior to send digital notifications to citizens, linking to all official publications relevant to their registered address.

Key features include:

1. Automatic inclusion for residents aged 25+ (opt-out model);
2. Weekly messages only;
3. Direct links to relevant notices;
4. Optional customization for topics or extra addresses.

The system operates via the national government's secure messaging platform, already in use for official correspondence. It connects tens of thousands of publications to millions of citizens weekly, in cooperation with 400 regional authorities. Low opt-out rates (≈7%) and minimal helpdesk demand indicate strong public acceptance.

This Dutch initiative illustrates how legal reform, technical innovation, and user-centered design can improve transparency, engagement, and trust in government. It offers a practical model for other jurisdictions seeking to make official information more accessible and personally relevant.

With a vast unmet need for legal services, large language model (LLM)-based legal question-answering systems offer unprecedented opportunities to bridge the justice gap by democratizing legal knowledge and providing clear, accessible, round-the-clock assistance. This paper examines their promise and challenges through two case studies: the CLIC Recommender (CRec) and CLIC-Chat. CRec is an LLM-powered AI recommendation tool that analyzes user queries and connects them to vetted Community Legal Information Centre (CLIC) resources. Its structured approach ensures reliability but is limited to pre-existing content. CLIC-Chat, by contrast, uses LLM generative capabilities to engage in interactive, conversational exchanges, clarify intent, and deliver detailed, context-aware explanations in plain language, enhancing accessibility.

However, public deployment raises legal and ethical concerns. LLMs risk crossing into unqualified legal advice, creating liability for inaccurate or outdated outputs, and lack professional oversight or indemnity protection. The issue of data privacy is equally pressing, as queries may contain sensitive personal information, demanding strong safeguards to comply with data protection laws. Balancing wider access with effective risk management is crucial. Access control models—from bans to supervised or fully open systems—illustrate trade-offs between accessibility and risk. Through CRec's structured approach and CLIC-Chat's evolving capabilities, this paper argues LLMs hold transformative potential but require careful navigation of legal and technical challenges to ensure trustworthy public use.

We'll discuss how Courthouse Libraries BC employed user research and our unique librarian expertise to revitalize the Clicklaw website, a trusted online resource for people looking for self-help guidance on their legal matters. Officially launched in October 2024, Clicklaw is one of the Library's digital services, powered by our law librarian expertise and real-life experience with public clients. The Library enhances public access to legal information through a user-friendly database and guides to common legal questions.

European legal scholarship has historically relied on commercial publishing, with a slow shift to open access. Recent policy frameworks – such as the 2021 UNESCO Recommendation on Open Science and the 2024 Barcelona Declaration on Open Research Information – have strengthened calls for open, transparent, and sustainable infrastructures. This study examines how legal scholarship is represented in OpenAlex, an open bibliographic database indexing over 268 million scholarly works, as a lens to assess the visibility of law outputs in open infrastructures.

We analysed 32,830 legal articles (2020–2024) with European author affiliations indexed in OpenAlex. Results show concentration in Russia (27.3%), the UK (11.4%), and Spain (10.5%). Open access prevalence is 62%, largely via Diamond OA (17.1%), but only 21% of OA works meet DOAJ criteria. English-language outputs dominate (65.4%), likely reflecting indexing bias. Metadata gaps remain significant: licensing is missing for 43.2% of OA works and persistent identifiers are inconsistently applied.

Italian outputs (n=1,240) mirror these trends, with 58.1% OA, mainly indexed via Crossref (76.4%). Cross-referencing with the most relevant Italian bibliographic legal database DoGi – Dottrina Giuridica, which covers 553 Italian law journals, shows that only 15 journals (2.7%) appear in OpenAlex, underscoring limited coverage of national legal literature.

Findings confirm the underrepresentation of legal scholarship in open infrastructures and stress the need for improved metadata quality, persistent identifier adoption, and integration of national resources to support a transparent and inclusive scholarly communication ecosystem.

The Canadian Legal Information Institute (CanLII) has evolved from a legal database into a comprehensive ecosystem that addresses access gaps through strategic partnerships and user-centered design. This presentation examines how coordinated initiatives, sustainable funding models, and collaborative content creation can effectively serve diverse user needs.

As Canada's largest free legal database, CanLII demonstrates a unique funding model that combines lawyer membership dues through the Federation of Law Societies with targeted foundation grants. This approach enables both comprehensive primary legal materials and secondary resources, including AI summarization tools that make complex legal information more accessible.

CanLII's Commentary program harnesses professional expertise for public benefit through the Authors Program. Legal professionals contribute commentary and analysis, while partnerships with publishers and academic institutions generate open-access materials with integrated deep linking. These collaborations create sustainable content development relationships that serve multiple stakeholder needs.

The recent platform redesign reflects CanLII's "no user left behind" approach, focusing on practical improvements to legal research accessibility. Enhanced search functionality and streamlined navigation support both legal professionals and self-represented litigants, while existing features like myCanLII, noting up, and legislative point in time continue to provide specialized research tools that reduce traditional barriers to legal information.

The Indigenous Law Project demonstrates how legal information systems can support reconciliation through respectful community engagement. Extensive consultation with Indigenous communities, legal scholars, and practitioners ensures increased access while maintaining cultural protocols and community autonomy.

This session will demonstrate how each initiative reinforces the others, creating a comprehensive approach to community-centered legal information access. Participants will learn practical strategies for creating community-centered tools that improve both information accessibility and user comprehension, providing actionable insights for Legal Information Institutes seeking strategic development.

Between 1927 and 1951, Section 141 of Canada's Indian Act prohibited the use of externally sourced funds to support legal claims against the Crown by Indigenous Peoples. While Section 141 has been cited as evidence of Canada's repressive approach to Indigenous advocacy, its concrete legal effects remain difficult to assess. To explore this historical dynamic, we examined the relationship between Section 141 and the legal record preserved in the Canadian Native Law Cases (CNLC), a major compilation of reported decisions involving Indigenous litigants, which was compiled in the 1980s and 1990s, and archival records from the period. A statistical analysis was applied to identify patterns of case reporting to see if they can be mapped to changes in litigation patterns arising from this law. This was then compared with extensive archival research carried out in multiple jurisdictions across Canada to identify the dynamics of how this law was enforced and as a comparator to see if the caselaw record reflected communities' experiences.

Understanding the relationship between reported case law and actual court activity is essential for interpreting historical legal trends, especially in contexts of colonial suppression such as the enforcement of Section 141 of the Indian Act. In common law systems, case reporting has long served as a mechanism for establishing precedent, but it is also a deeply curated and partial record of legal activity. The CNLC, while a resource for examining the history of Aboriginal law, must therefore be read with a critical eye toward its archival and institutional limitations. As with all law reporters, the CNLC must be read not as a transparent reflection of judicial reality but as a curated legal artifact, shaped by editorial discretion, institutional bias, and normative expectations of legal form.

The Towards Truth project, a collaboration between the Justice and Equity Centre (JEC) and the UNSW Indigenous Law Centre (ILC), is a first of its kind tool mapping in detail the vast body of laws and policies that have impacted Aboriginal people since 1788. The project is a response to the Uluru Statement from the Heart's call for Truth, and it shows the importance of innovative ways to deliver access to justice and provide legal information to communities through online platforms.

This presentation will showcase the Towards Truth website, which explains, in clear, concise language, relevant laws and policies and their impacts on Aboriginal people in NSW. It will also explain the multiple potential uses of Towards Truth – first and foremost to support local and grassroots truth-telling efforts, but also as a resource to be used in schools, universities, workplaces and by the general public to increase understanding about the role and impact of laws and policies on First Nations people.

This presentation will show how Towards Truth can play a vital part in truth-telling in Australia, not only by documenting historical injustices but also by celebrating the resilience and cultural endurance of Indigenous communities. In doing so, we seek to expand the understanding of the role accessible, online legal information can play and demonstrate how innovative resources can serve as a catalyst for nationwide examination of past injustices and their contemporary reverberations, laying the groundwork for meaningful reconciliation and social change.

The unauthorised extraction of digital records of First Nations' cultural heritage from open online collections is being used to train Artificial Intelligence (‘AI’) systems. These practices risk undermining Indigenous data sovereignty, stripping cultural expressions of context and protocols, and enabling their reproduction or manipulation in ways that are disrespectful, cause harm, and are misleading. The question arises, how can GLAM institutions (Galleries, Libraries, Archives, and Museums) that create and/or hold such digital repositories serve as trustworthy stewards of First Nations' knowledge in an era of generative AI?

This research builds on the first author's previous work on developing legal and governance frameworks for the management of First Nations' knowledge and cultural expressions, and seeks to extend this work to the digital infrastructures of GLAM institutions. Rather than advocating for reduced access or digitisation, the research aims to support institutions in designing legally sound systems that are both culturally sensitive and technologically resilient. The goal is to embed established ‘Indigenous governance principles’ into metadata structures, access protocols, and rights management – ensuring that trust is built into the very architecture of digital cultural collections.

Complementing government endeavours to protect First Nations arts, crafts, sciences, technologies, and cultural heritage, the aim is to inform and contribute to policy developments and best practices related to digital trust and safety in this sphere. It builds on foundational governance research conducted through the Garuwanga Project (led by the first author) and adjacent work on First Nations cultural heritage law, knowledge management, and ethical digitisation.

The legal profession's AI divide is not between adopters and sceptics, but between those who understand the underlying technology and those operating without technical insight. This presentation examines the engineering principles that determine whether AI systems enhance or compromise legal practice.

We will begin with the topic of semantic search, a fundamental shift from keyword matching to conceptual understanding. Through technical demonstration, we explore how vector embeddings transform legal language into mathematical representations, enabling systems to recognize conceptual relationships across varied terminology. This allows natural language queries to surface relevant authorities that Boolean searches miss, representing a paradigm shift in legal information retrieval.

Retrieval-Augmented Generation (RAG) addresses AI reliability challenges. Unlike Large Language Models that generate statistically plausible text, RAG architectures constrain responses to verifiable legal sources. We examine how these systems retrieve documents, construct context windows, and generate responses with traceable citations. Importantly, we analyse why implementation quality varies significantly across platforms, with some achieving sub-paragraph citation precision while others provide only superficial source attribution.

The presentation provides systematic analysis of hallucination phenomena in legal AI. We demonstrate how transformer architectures generate fictitious but plausible legal information, examine conditions triggering such failures, and present verification protocols for detecting phantom citations. International examples illustrate both failure patterns and successful mitigation strategies.

Our technical evaluation of current market offerings reveals substantial disparities between marketing claims and architectural capabilities. Through comparative analysis of leading platforms serving the Australian market, we distinguish genuine RAG implementations from prompt engineering overlays.

Finally, we explore emerging AI agents capable of autonomous multi-step legal research. While offering transformative potential, they introduce compounded risk vectors requiring careful evaluation. We examine their architecture, capabilities, and failure modes.

Attendees will gain practical methodologies for evaluating AI tools, implementing verification protocols, and contributing to institutional technology strategy with technical confidence.

Discover how Oracle’s AI Strategy is helping organisations deliver business outcomes and innovation. By leveraging curated frontier models,AI infrastructure and embedded AI services,organisations are accelerating the build, deployment and adoption of new AIpowered services —delivered with governance, privacy, and security by design.

The unauthorized use of copyrighted content in AI training has sparked widespread concern and over 40 lawsuits in the US. While legal scholarship has focused on copyright applicability and fair use, less attention has been paid to a critical issue: transparency around AI training data. Without knowing whether and how their content is used, right holders cannot enforce or license their rights. Although policy proposals in the US, EU, and UK aim to improve data transparency, their effectiveness remains unclear. This article explores the meaning and necessity of AI data transparency in copyright and broader contexts. It critically assesses current and proposed measures—such as dataset disclosures, data summaries, and certifications—highlighting challenges including technical feasibility, legal risks, and trade secret protection. It then proposes a flexible, accountable, and holistic framework for AI data transparency that balances stakeholder interests and addresses existing limitations in transparency mechanisms.

The ‘free access dilemma’ arises from wanting to make data free for individual access via the Internet, but not for bulk downloading, including for AI training sets. Put briefly, ‘how to allow access but retain protection?’ The dilemma applies to many Internet publishers, including GLAM institutions (Galleries, Libraries, Archives and Museums), and also to some LIIs (Legal Information Institutes). The dilemma can apply to both content which is copyright, and content which is (technically) in the public domain. LIIs are the focus of this paper. In Australia, the four existing methods of protecting data are contracts and licences, copyright law, the Robots Exclusion Protocol (REP) and blocking robots. Nothing works well enough. What reforms are possible? Apply foreign laws (‘Internet embassies’)?; A fair dealing exception (Legalise theft); A compulsory licence (Mandate arbitration) – possibly with opt-out; or Reform the REP (Payment options). A good policy conclusion is hard to find.

As AI systems become ubiquitous and increasingly opaque, it is difficult for regulators to effectively monitor their effect, identify unlawful AI-enabled and AI-assisted actions and collect the evidence needed to undertake legal proceedings. In a society shaped and transformed by AI, a new paradigm for law enforcement is necessary. Designing and enacting AI laws is the easy part. Ensuring that such laws can be effectively enforced is the greater challenge. While the European Union has enacted an AI Act and governments around the world have released proposals for AI legislation, there has been limited consideration of how such laws can be enforced. The purpose of this presentation is to suggest a new ‘compliance by design’ approach. This involves embedding technical controls, protocols and tools within the design of AI systems to support legal compliance. This is an intrinsically interdisciplinary exercise, requiring the collaboration of lawyers, computer scientists, domain experts and others. A wide spectrum of laws, including data privacy, consumer, corporations, administrative, and anti-discrimination laws, have the potential, to differing degrees, to be embedded in this way. It is suggested that such a strategy of compliance by design offers a logistically feasible and technical solution to the critical legal problem of our time – the safe and responsible deployment of AI.

The widespread availability of generative AI technology has shifted the operating environment for legal publishers more profoundly than any other shift since the advent of the World Wide Web. We've known we were running a Red Queen's race with regard to feature development in an environment of rapid technological advancement. On the other hand, we'd come to expect that data and technology would mostly only improve.

As we shift from an environment in which our data was outside most models to one in which our data was inside most models to one in which our data was used to generate similar content that is now also inside those models, our ability to maintain clear end-to-end understanding is hampered. This document describes some of the approaches LII has taken to overcome the challenges we currently face.

LII states its mission as: "We explore, prototype, refine, implement, and share innovative techniques for making legal information more findable and understandable for the public. We value accuracy and objectivity, both in the technologies we use and the content we publish." In other words, we use technology to help people find and understand the law. Our focus on people most centrally enables us to frame our experimentation and build thoughtfully on a foundation of what we already know.

This presentation will detail examples of shifts in technology and information availability. It will detail examples from recent collaborative projects of the LII: the classification of regulatory purpose documents, which surfaced the "why" or "what for" behind regulatory texts; the automated representation of regulatory text using the L4 domain specific language, which aimed to make public benefits eligibility more efficient to determine; ontology linking, which connected state-level regulations to the SALI/SOLI/FOLIO ontology, bridging rules with practice; and finally, the combination of techniques behind search enhancements enabled LII to select the best tools to use for different types of exploration. The presentation concludes with gleanings from this work for our public service mission.

This presentation outlines AustLII's implementation of Retrieval-Augmented Generation (RAG), an AI architecture that enhances large language models (LLMs) by integrating them with real-time retrieval from AustLII's legal databases. RAG addresses limitations of standard LLMs, such as outdated information, lack of source attribution, and risk of citation errors, by retrieving current, jurisdiction-specific legal materials. Using a combination of semantic and keyword-based search, the system identifies relevant documents and synthesises them into structured responses with appropriate citation. This approach supports more reliable legal research and illustrates how RAG can be applied to improve free access to legal information.

This presentation offers a case study of Lexum's initiative to integrate large language models (LLMs) into its publishing workflows by adding AI-generated enrichments to legal documents published on CanLII, Canada's open-access legal information platform.

After a brief recap of the project's objectives and evolution, the session will outline Lexum's current progress in automating content enhancements while maintaining strict validation safeguards and improving workflow efficiency to deliver more accessible and user-friendly legal information.

The presentation will also examine the unique challenges of deploying generative AI in this context, including Canada's bijural and bilingual legal system, the rapid pace of AI development, stakeholder apprehension around reliability and transparency, and growing back-end complexity.

Grounded in a real-world implementation, this case study offers a practical perspective on how generative AI can be leveraged to enhance legal research tools and improve access to justice in Canada's digital legal landscape.

AI offers new and novel ways of achieving successful pedagogy in higher education. However, the abilities of AI also have the potential to undermine the intellectual and academic development of students. In particular, the hyper-efficient capabilities of AI disrupt the traditional and accepted paths to developing expertise. Using law and legal education as an example, I will explore the problems and possibilities that AI offers in the field of tertiary education. The problem of a wise algorithm lies in the possibility that some AI technologies may develop capacities that outstrip that of both learner and teacher. The replacement problem posed by AI exists on multiple levels and calls for a deeper debate on how the technology can develop in a way that strengthens universities rather than posing yet another challenge to the beleaguered sector.

This presentation introduces AI Moot Court, a novel multi-agent system employing large language models (LLMs) and retrieval-augmented generation (RAG) to simulate appellate advocacy for legal education. Traditional moot court training requires significant resources, including dedicated venues, extensive personnel, and limited opportunities for student practice. AI Moot Court addresses these constraints by offering an interactive, scalable, and procedurally faithful virtual environment accessible to students remotely.

The system enables users to take on the roles of appellant or respondent, interacting with AI agents who simulate the judge and opposing counsel. Using a structured sequence that mimics real appellate hearings—such as skeleton submissions, oral arguments, judicial questioning, and rebuttals—the platform fosters practical advocacy skills. Integrating RAG with Hong Kong's legal corpus from the HKLII database ensures that generated arguments are grounded in authentic case law, enhancing legal accuracy and mitigating hallucinations common in LLM outputs.

The judge agent evaluates participant performance via the IRAC framework (Issue, Rule, Application, Conclusion), providing detailed feedback and scoring on legal reasoning, precedent application, fact analysis, and advocacy proficiency. Multiple LLM models were evaluated, with multi-agent coordination and RAG shown to improve coherence and factual grounding.

By blending AI capabilities with established legal pedagogy, AI Moot Court expands access to robust advocacy training, offering a realistic, interactive supplement to traditional moot courts. This work demonstrates the potential of generative AI to transform legal education through innovative, technology-driven solutions.

The UTS Law Tech Challenge consistently ranks as students' favourite experience while at university. The Challenge is also a trailblazing example of how a motley gathering of people and organisations can improve access to justice. ‘The most rewarding experience I've had throughout university,’ is one student's description. ‘It's rare that I can have an opportunity to apply my skills towards making a difference in something real, and more meaningful.’ Since 2016, the Challenge has enabled teams of law students to build apps for not-for-profit clients, aided by law mentors and tech experts. The ARISE Foundation app from 2024 helped victims of domestic violence re-enter the workforce; an app for the International Committee of the Red Cross educated users about international law; an app for Sydney Story Factory taught creative kids about defamation and copyright; and a series of community legal centre apps streamlined intake processes to enable thousands more efficient access to low cost legal services.

In this paper, I give the lowdown. That includes detailing the role of Wotton Kearney, who sponsor the Challenge, and who provide legal mentors for each student team, as well as specific legal expertise. It also includes explaining the role of software company Neota Logic, who play a hands-on role in building apps, and in preparing students for a tech-rich future. And the Challenge throws up challenges, of course, including confused communications and student frictions. The aim, ultimately, is to explore what the Challenge can teach us about improving access to justice, with the hope of inspiring like-minded initiatives. As one not-for-profit client said, ‘This collaboration has been a standout example of what's possible when academic enquiry, private sector expertise and youth-led innovation intersect with purpose.’

For rigorous data-driven research in the analysis of judicial practice focused on a specific legal domain or legal institute, the first step is to obtain a complete sample of court decisions relevant to the area of interest. We present a guided, semi-automated methodology for creating annotated datasets from legal corpora, tailored to specific legal domains.

The process starts with the dataset creator specifying key legal provisions, phrases or a free-text description of the targeted legal domain relevant to the area, or their combination, to identify a superset of potentially relevant court decisions. The result is several sets of documents, likely containing a large number of irrelevant ones.

As an example for demonstrating our methodology, we focus on the area of cybercrime. In the case of cybercrime, we used the Convention on Cybercrime as a textual input. Based on the semantic similarity between this text, the content of court decisions, and the documents they cite, we generate an initial estimate of the relevance of candidate documents.

The next step is semi-automated. The user manually evaluates the relevance of several documents and classifies them as relevant or irrelevant. By analyzing the annotated documents, the system gradually learns to classify documents with increasing accuracy. In addition, the system suggests documents that do not match the initial query but are semantically similar to what the user is looking for. This is done to uncover documents the user may not have explicitly searched for, but would still want to include in the dataset, thus making the final dataset as complete as possible.

Using a human-in-the-loop approach, the process continues until the user feels that the system classifies with sufficient accuracy and obtains a final set of documents for further legal analysis.

The process can continue with the extraction of various document features and subsequent analytical tasks such as clustering, visualization, frequency analysis of legal concepts, or outlier detection.

Access to justice remains a significant challenge for the indigenous communities living in the outer islands of Vanuatu. Geographical isolation, limited legal resources, and a lack of timely and accurate information often hinder informed decision-making and delay fair judicial processes. In response, this project is developed in partnership with the Judiciary of the Republic of Vanuatu and supported by USP/PacLII, and it explores how technology can support and improve the preservation and accessibility of legal information, particularly from lower courts and customary court systems.

Traditional methods of preserving legal records have proven vulnerable to natural disasters, climate change, and resource constraints. As a result, vital judicial information stemming especially from village, local, and customary courts has been lost. This project proposes a digital approach to ensure the long-term preservation of legal records, enabling a more accurate and complete reflection of the legal process, from lower courts through to appellate courts.

The project places strong emphasis on involving community leaders, chiefs, judicial officers, and local court staff in the digital preservation of legal records and in building their legal knowledge through practical training. By providing targeted training in legal information, case management, and digital preservation, the initiative aims to reduce delays in the publication of land disputes, chiefly titles, and customary matters which are issues of deep cultural and legal significance in Vanuatu.

More than a technical exercise, this project recognises the importance of cultural sensitivity and the need to respect Vanuatu's diverse customary laws while improving legal transparency and accessibility. Early outcomes show promising improvements in court efficiency and information flow, demonstrating how appropriate use of technology can bridge systemic gaps in justice delivery.

Ultimately, the project aims to empower communities in remote areas with fairer, faster, and more transparent access to legal information and services and in strengthening both the formal and customary justice systems in ways that are sustainable and locally grounded.

This presentation introduces a new AI-assisted search feature developed by Lexum for the CanLII platform. The tool is designed to improve both the formulation of legal queries and the interpretation of search results using Large Language Models (LLMs), while remaining grounded in principles of transparency, control, and accessibility.

The feature assists users at two key stages of the legal research process. First, it interprets natural language questions and converts them into precise legal search queries. Second, it synthesizes and presents search results in a clear and readable format, ensuring that the original sources remain visible and accessible. Unlike black-box AI interfaces, this system prioritizes explainability: users retain full control over the dataset being queried and are always able to see how AI-generated outputs are produced.

The presentation will provide an overview of the system's development, including user validation feedback, funding context, and the design decisions informed by Lexum's previous work with AI-generated legal content. A live demonstration of the search feature will illustrate how it fits into CanLII's mission of delivering high-quality, publicly accessible legal information.

With a full rollout planned for 2026, this project represents a pivotal step in Lexum's evolution—leveraging AI to make legal research more intuitive and inclusive, while upholding the core values of transparency and trust. This session offers practical insights into the responsible deployment of generative AI in legal information environments.

Scholars and practitioners in law and legal information can democratize knowledge, accelerate legal information access, and engage in shared legal collaboration by fostering public access to their work. Legal scholars and information practitioners may use open licenses such as under the Creative Commons framework, distribute their work with disclaimed copyright, or simply share their work in open access journals, institutional repositories, and open-access legal platforms. Open access policies from agencies such as Canada's Tri-Agency Open Access Policy and access to justice ideals encourage or even require such practices.

However, the rise of artificial intelligence (AI) and large language models (LLMs) presents a challenge: Openly accessible knowledge intended for scholarly exchange and legal information distribution and access may be enveloped into massive training corpora without scholars' or practitioners' consent or awareness, or under licenses previously granted when such use was not contemplated. Commercially developed LLMs can be sold as products to legal scholars whose very expertise, knowledge, and access to justice spirit formed the expert corpora from which the legal content, data structures, and legal metadata was generated.

In effect, knowledge built by legal scholars and practitioners through research, human expertise, and an access to justice ethos is transformed—appropriated—into data that drives models. The data-information-knowledge cycle is inverted, with human knowledge datafied and exploited by commercial AI and LLM developers and providers of commercial legal information intermediaries. Many of these are controlled by interests centred in the global north, in large commercial players, or both. Greater digital colonialism, knowledge datafication, and environmental waste exacerbate global disparities, differential power structures, and differential access to legal information, and their corresponding influence on knowledge growth and direction and legal information access. All of this defeats global knowledge democratization and access to justice goals.

We explore counter-strategies at the levels of individual scholar and legal information practitioner, research institutions, and governmental or agency frameworks: Scholars and legal information practitioners may limit use of their knowledge in LLMs and other AI models using negotiated licenses, asserting Creative Commons license modifications, and employing intellectual and other private law mechanisms to prohibit (or seek compensation for) knowledge appropriation. Institutions can prioritize legal scholarship and information safeguards, while furthering the ethos of open access, global equity, knowledge democratization, access to justice, and digital decolonization.

Institutions can constructively exploit expertise through both internal collaborations among scholars and practitioners of science and technology, law and policy, and libraries and institutional repositories, and inter-institutional collaborations in the global south and north. At the governmental or framework level, initiatives include regulation, legislative revision, and advocacy for updated governance and frameworks to further the goals of access to justice and disparity reduction.

Ultimately, we explore solutions to sustain the digital legal knowledge and information commons, ensure continued value of human legal expertise, and reduce global inequities in legal knowledge access and control, while supporting individual, institutional, and societal innovation through responsible and ethical use of evolving AI technologies that can indeed advance access to justice.

The term “privacy” is often contrasted with “publicity.” Many assume personal information loses its private nature once disclosed in public spaces, such as through court judgments or administrative actions. However, advanced technologies, particularly artificial intelligence, challenge this view. Private entities can analyse publicly disclosed legal documents to uncover latent details, creating inferred information like judicial profiles or social credit reports of individuals. While disclosed information is presumed to forfeit privacy rights, inferred information, unanticipated by data subjects, raises ongoing debates about privacy retention.

Using sophisticated tools, private entities extract and repurpose personal information from legal databases, generating information products for profit. This practice, while resource-efficient, risks informational privacy by relying on vast personal data from legal sources. Not all inferences demand protection, nor do all activities require stringent oversight. Balancing economic benefits and legal transparency with privacy risks is critical.

This article explores promoting technological innovation through analysing publicly disclosed personal information while safeguarding data subjects' informational privacy. First, it examines data mining in court judgments and official documents using AI, arguing that inferred content retains privacy. Second, it analyses the value orientation between technological innovation and privacy protection by comparing China's regulatory framework with other jurisdictions. China's framework often deviates from its value orientation, creating practical dilemmas. Finally, it proposes an enhanced regulatory approach, placing greater responsibility on private entities to protect privacy while easing burdens on data subjects. This framework aims to align innovation with robust privacy safeguards, ensuring compliance and transparency without stifling technological progress.

Forthcoming

The new capability of AI, particularly GenAI, provides an opportunity to change how legal assistance is provided to legal-help seekers with potential to improve access to justice. There are also risks associated with the use of these new technologies, including that the use of AI: may provide inaccurate or incomplete legal assistance to a legal help-seeker; may increase the work-load of the courts compelled to respond to poor quality legal analysis; and may contribute to the spread of legal misinformation and disinformation undermining social cohesion and the rule of law.

These potential benefits and risks of AI raise questions about the suitability of the current regulatory model—the laws relating to the regulation of lawyers and legal services—when applied to the provision of legal assistance through AI. This talk will address those issues. It will consider firstly whether the existing regulatory model that focuses on regulating authorised legal service providers is fit to deal with the new challenges raised by the use of AI to provide legal assistance. It will argue that the law in its current state is not fit for purpose and there is an urgent need for reform.

It will consider secondly, what kind of reform is required to ensure that if AI is used to provide legal assistance that the fundamental legal services values (loyalty, competence and confidentiality) are upheld, the institution of the law is protected, and AI is, in fact, used to address the access to justice gap.

This paper examines an AI-native law firm model in Ontario designed to address the access to justice crisis facing self-represented litigants. The model integrates AI technologies, including document generation, legal research, note-taking, and courtroom simulations, all with a layer of human oversight. It demonstrates how AI can expand access to justice while maintaining professional quality and compliance with the Law Society of Ontario's regulations.

The paper also explores the responsibilities that legal practitioners, judicial systems, and self-represented litigants hold in balancing AI risks against benefits, arguing that regulated AI adoption represents a viable path towards improving access and reducing barriers to legal services.

Katherine Hayles and Donna Haraway's posthumanist theory is a transformative paradigm that challenges the traditional notion of humans as stable, independent entities. Instead, they propose a dynamic perspective of humans as relational beings who are constantly influenced by interactions with politics, technology, the environment, and history. Hayles envisions the posthuman subject as a composite entity, fluid and diverse in its components. The cyborg, a concept that blurs the lines between human and nonhuman, organic and artificial, is introduced by Haraway. Our perceptions are challenged and our curiosity piqued by the existence of this machine-organ hybrid in both fiction and social reality. By rejecting the idea of humans as unified, autonomous, and self-contained, posthumanist thought necessitates a reevaluation of legal frameworks, which emphasises that technology is more than just a tool for the outside world; rather, it is a part of the ever-evolving relationships between humans, animals, and machines.

This perspective presents humans as relational; hybrid beings co-produced with technological systems. If humans are already entangled with the nonhuman world ecologically, emotionally, and technologically, then legal frameworks based on rigid human-centred boundaries become inadequate. Legal systems treat technology as a neutral instrument, but posthumanism challenges this view. These systems still assume a human-nonhuman divide that posthumanism has questioned. The blurring of boundaries in the posthuman world necessitates a rethinking of legal frameworks. The current frameworks are ill-equipped to handle the complexity and interconnectedness of this new reality. The posthumanist legal subject demands a radical reimagining of legal information in its content, form, function, and knowledge base.

In what may soon cease to be speculative fiction, one might imagine a reality where legal information becomes a biocomputational interface, embedded in neural chips, synthetic minds, and planetary governance systems. Normative boundaries are sensed, cognitively internalised, and enacted by the posthuman subject to adapt their behaviour. Such boundaries are what we refer to as 'embedded legal consciousness', a concept that suggests that the subject (the posthuman in this regard) has internalised law, turning it from external regulation into behavioural cognition, which is precisely where H.L.A. Hart's concept of the "internal point of view" acquires renewed relevance. By this understanding, posthuman subjects do not merely follow or consult rules externally (from fear or coercion); they inhabit and adapt them through embedded legal-awareness modules, demonstrating functional acceptance.

In posthuman jurisprudence, agency, normativity, and interpretation are no longer confined to human realms. Legal reasoning transcends 'human' cognition, becoming a function of distributed, adaptive intelligences. Law is no longer restricted to text but is embedded in code, sensors, and networks, executed and interpreted in real time by recursive, affective systems. Legal judgment does not stem from individual human deliberation, but from collective, post-anthropocentric cognition.

This position aligns with Dworkin's framework of judges as epistemic agents, but extends it to nonhuman, composite, and hybrid intelligences. This shift redefines law as a co-evolving epistemic architecture, where meaning is modulated across hybrid posthuman agents. Recognising the fluidity of personhood expands the scope of legal information, allowing it to encompass emerging techno-social actors. Legal precedent evolves as a dynamically reweighted structure, responsive to ecological, social, and algorithmic landscapes.