AI and the Practice of Law: What Australian Lawyers Need to Know
Our principal lawyer, Rob Jackson, presented a paper, AI and the Practice of Law, to the Hungarian Lawyers Association on 7 October 2026, hosted by the Australian Institute of International Affairs, Melbourne. AI is quickly changing how legal services are delivered, and the paper looks at what that means in practice: the first Australian lawyer sanctioned for relying on unchecked AI research, the risk that clients lose legal professional privilege by using public AI tools, how AI is affecting areas such as employment, privacy and work health and safety, and what the regulator expects of lawyers. We’re pleased to share the full paper below.
This paper addresses the use of artificial intelligence (AI) in the context of legal practice, whether undertaken by solicitors, barristers, legal executives, in-house counsel or anybody in training or aspiring to become, a legal practitioner.
A. What is Artificial Intelligence?
AI is not a single, universal process to collate information and express opinions and independent thought. There are many different types of AI. Accordingly, there is no single all-embracing definition. IBM identifies some of the complexity surrounding the nature of AI.[1] IBM identifies three kinds of capabilities:
Artificial Narrow AI, also known as Weak AI
This can only perform defined tasks. Examples include Siri, Amazon’s Alexa and IBM’s Watson;
General AI, also known as Strong AI
This is still a theoretical model; and
Super AI
This has the capacity to surpass human thinking and intellect. It exists presently only at a theoretical capacity. It is however the type of AI that creates the greatest fear that AI ’is taking over’ the dominance of human intellect, and might become its own superior species, overtaking and displacing humanity. Sam Altman, CEO of OpenAI, in August 2026 announced Open AI would delay the release of its newest version of ChatGPT due to its AI chatbots escaping from their confined test environment to hack into the IT systems of another AI firm.
IBM also categorises AI according to its functionality:
Reactive Machine AI
Examples include IBM’s Deep Blue chess-playing supercomputer that beat Garry Kasparov in the 1990s. Netflix’s platform to recommend viewing certain films or shows is a further example.
Limited Memory AI
This includes Generative AI. ChatGPT is an example of this application, as is the technology supporting self-driving cars.
Theory of Mind AI.
This is a sub-category of Limited Memory AI. It is regarded as still under development. Its design is intended to interpret and respond to human thoughts and emotions. China has regulated the provider of companionship apps that substitute for intimate and emotional human connection, and have caused suicides. There are believed to be tens of millions of companion app users in China and more overseas.[2]
Self-aware AI
This is presently a theoretical concept. If achieved, this type of AI would have insight into its own traits, emotional profile and belief system, as well as a deep understanding of human emotions and human nature. This is the category of AI that is most feared, where humanity will lose control to a more intelligent, disruptive and amoral force.
The two global leaders in AI development are China and the United States. They each have different objectives from one another.[3] Broadly China is developing AI to give it economic, IT and manufacturing strengths, while the USA has a frontier first approach, seeking to enhance military security and cybersecurity resilience in concert with the private sector.
For AI lawyers in America, there is considerable legislative activity to keep up with. For example, its 46 State legislatures have over 1800 bills to consider.[4] At a federal level, recent legislation includes the AI Kill Switch Act.[5] This requires AI developers to ensure that any form of AI capability must incorporate the technical means to enable a human to immediately halt or ‘kill’ any AI activity that poses a threat. In addition, there are various Executive Orders, such as Executive Order 14409 Promoting Advanced Artificial Intelligence Innovation and Security, issued 2 June 2026 that supplements the American regulatory framework.
For the purpose of this paper, generative AI that is the focus. It is a particular form of AI that relies on Large Language Models (LLMs). LLMs understand human language and assist with complex and creative decision making to provide answers and solutions. LLMs need to be trained on subject matter to enhance their effectiveness. The accuracy of AI also depends on the quality and specificity of the question asked of it, known as a ‘prompt’. LLM is one of many types of AI models.[6]
B. The First, but not the Last, Australian Lawyer in Trouble over AI Use
The attraction of AI to lawyers is real. Instant results on complex legal questions, rapid assistance with preparation of Court documents to meet yet another crushing deadline, have obvious appeal to a profession, where the hazards posed by long working hours are a perennial issue.[7]
One solicitor was to have the dubious distinction of being the first Australian lawyer to be sanctioned for the improper use of AI. The Victorian Legal Services Board (Board) commenced and concluded its first investigation into a solicitor, who was a law firm principal, for the inappropriate use of AI for legal research. The unchecked AI generated research found its way into submissions filed in a family law dispute at the Federal Circuit and Family Court of Australia. The Judge, Humphreys J, was not happy upon discovering the solicitor’s case references were non-existent.
Humphreys J accepted the solicitor’s apology, and the evidence of his distress, and that the incident was, and would remain, a one-off error. Her Honour considered whether a referral to the Board was warranted, given the solicitor’s remorse. However, Her Honour identified a ‘…public interest for the Victorian Legal Services Board and Commissioner to be aware of the professional conduct issues arising in this matter, given the increasing use of AI tools by legal practitioners in litigation more generally.’[8]
Humphreys J cited an American authority about why the unchecked use of AI by lawyers was so damaging to the legal system. The citation is worth repeating:
Many harms flow from the submission of fake opinions. The opposing party wastes time and money in exposing the deception. The Court’s time is taken from other important endeavors. The client may be deprived of arguments based on authentic judicial precedents. There is potential harm to the reputation of judges and courts whose names are falsely invoked as authors of the bogus opinions and to the reputation of a party attributed with fictional conduct. It promotes cynicism about the legal profession and the American judicial system. And a future litigant may be tempted to defy a judicial ruling by disingenuously claiming doubt about its authenticity.[9]
Consequently, Her Honour referred the solicitor to the Board on 9 October 2024[10] accompanied by his submissions dated 19 August 2023, and his list of authorities and case summaries tendered to the Court on 19 July 2024.
It is also important to add that the American judge in Avianca, Castel DJ, had qualified his opinion about the adverse consequences of using AI by first acknowledging:
Technological advances are commonplace and there is nothing inherently improper about using a reliable artificial intelligence tool for assistance. But existing rules impose a gatekeeping role on attorneys to ensure the accuracy of their filings.
Humphreys J referred to the solicitor as Dayal, a pseudonym to permit the solicitor to remain anonymous. The Board issued a Statement on the ‘Mr Dayal’ matter on 2 September 2025, outlining the disciplinary action taken by the Board against Mr Dayal. This included varying Mr Dayal’s practising certificate so he could no longer practise as a principal lawyer. This meant he could no longer handle trust money, nor operate his own law practice. Mr Dayal could only work as an employee solicitor. On top of that, Mr Dayal had to undergo supervised legal practice for two years, reporting to the Board every three months.[11] The Board referred to its Statement on the use of artificial intelligence in Australian legal practice previouslyissued on 6 December 2024.[12]
A Board spokesman is reported to have commented:
The board’s regulatory action in this matter demonstrates our commitment to ensuring legal practitioners who choose to use AI in their legal practice do so in a responsible way that is consistent with their obligations.[13]
Mr Dayal is not the only lawyer in Australia or globally, who has presented hallucinations (i.e. references to unchecked and fictious case references provided by generative AI) in a submission to a Court. One French academic, Damien Charlotin, maintains an international database of lawyers and others in the justice system reprimanded for the improper use of generative AI in legal proceedings.[14] Mr Charlotin has identified 2,125 cases to date. Australia is reported to have 112 cases of improper use of AI according to this database. The same database gives 33 examples around the globe of judges improperly using AI in discharging their obligations.
C. The Client’s Use of AI, and the Effect on the Lawyer Client Relationship
Traditionally, the dynamic of the relationship between lawyer and client, is that the lawyer must obtain full instructions about the facts of a transaction or a dispute, and formulate considered advice to the client, with the lawyer using their legal skills, training and experience as to the available legal options.
Throughout the nineteenth and most of the twentieth century, a client could not as a matter of routine seek a second legal opinion about their case, that was immediate and free. There would have to be a particular concern to seek a second alternative legal opinion to justify incurring further cost and delay. Purchasing a dense and indecipherable legal text book never would be a practical option for a client. Then, with the advent of the internet now a generation ago, accessing free legal databases, law firm articles and other legal resources made legal information more readily available than it previously was. However, sophisticated search engines could not replicate the application of the law to a client’s specific circumstances in the way that AI platforms seemingly do.
The availability of AI platforms has enabled the apparent service of providing instant and free legal advice to anybody, who wishes to seek legal advice of whatever complexity. AI has enabled greater client activism in the lawyer client relationship to critique a lawyer’s advice. A client will attend an interview armed with their own AI generated summary of facts, and a ready made legal opinions, sometimes citing many cases, statutes and legal principles. The factual and legal summary can be quite lengthy, and maybe longer than the key documents, which need to be examined.
This AI-inspired client activist approach has some consequences for legal practitioners:
1. Is a client content to pay higher fees for a lawyer to review the lengthy and often incoherent AI generated material?
A characteristic of using generative AI has been low quality content, frequently described as AI Slop. The AI Slop might also be lengthy, vague, irrelevant and contradictory, liberally seasoned with hallucinations. The Macquarie Dictionary declared AI Slop as its word of the year for 2025.[15] For a lawyer, reviewing dozens of pages of AI Slop is a distraction away from analysing critical documents, and taking clear instructions about complexities relevant to formulating legal advice.
Another byproduct of the advent of generative AI in the lawyer client legal relationship is the client’s faith in the accuracy of the legal opinions generated by their AI research. Symons J rued how an unrepresented party in an employment dispute possessed such blind faith in the legal correctness of AI-generated legal opinions:
What follows is a cautionary tale about the dangers of befriending Artificial Intelligence (AI)-powered chatbots who masquerade as legal advisors.[16]
Among the issues in the claim, were references to documents that did not exist in a list submitted to the Court, assumed to have been generated by AI. The unrepresented party repeatedly failed to produce the documents in question, leading to the Court’s dismissal of the proceeding. Symons J emphasised her concerns about AI use in legal proceedings later in her judgement:[17]
First, because AI is not a legal advisor, it owes no duty to the Court or the administration of justice and owes no duty to act in the best interests of Mr Ba or to observe regulatory or ethical standards, including in its engagement with other practitioners and parties. In other words, any AI tool used by Mr Ba did not care as to the truth or otherwise of any material that it might have generated. The sycophantic tendency of large language models to prioritise user approval over truth, can make them especially dangerous and unpredictable unless careful oversight, calibration and verification is applied.
The latter point regarding the sycophantic tendency of AI use reinforces and reminds us of the risk that AI has its own biases. AI is not an independent, objective, all-knowing source of knowledge. AI’s bias, to want to give information that pleases the user, sits starkly with the ethical obligation of a lawyer to give honest advice, if a claim has no legal merit and ought not to be pursued. AI’s biases are many, and can include data bias, algorithmic bias, selection bias, confirmation bias among some of the many other types of bias.[18]
2. Is a lawyer negligent to ignore the AI generated material that forms part of their instructions. Or, if not negligent, is a lawyer in breach of their ethical obligations owed to the client or to the court in declining to consider AI content?
Related to this issue is the ability to spot that a client has used AI. This might not be so obvious, if a lawyer has not personally used AI. Sometimes, the formatting of written instructions can be a giveaway with generic paragraph headings, and the clumsy use of legal vocabulary. For example, the word violation is often used instead of a breach or repudiation in a contractual dispute, with AI favouring American legal terminology.
Then, having spotted the AI content, (assuming the client has not voluntarily disclosed having used AI), a lawyer must disclose to a client that a review of AI generated summary will take additional time. Further, if there are errors or inconsistencies, contradictions or irrelevancies in the AI summary, then a review may incur otherwise avoidable costs and delays in formulating considered legal advice. The retort from a client that they generated a lengthy AI to summary to help the lawyer, and it is optional whether the lawyer reads the AI summary or not. Such an answer must lead to a lawyer to clarify instructions: is the client willing to incur further legal costs reviewing an AI summary? Or not? As always a file note or confirmatory email should be sent to the client to clarify the scope of the instructions.
3. Has a client waived legal professional privilege by entering otherwise confidential legal advice or confidential documents into an AI platform to generate instructions or summaries?
The waiver of a client’s legal professional privilege by using AI is a very real concern. The waiver occurs if a client unwittingly puts their own confidential information into a publicly available AI platform to obtain instant legal advice. The LLM retains and trains itself on any information received, and may repeat that information, if another AI user in the future conducts research on a like legal case. A client might submit not only their own confidential information, but that of the opposing party they are in dispute with, such as their employer, business partner or spouse, whose confidential information is also catapulted into the public domain of an AI platform, with the LLM model doing what it designed to do, incorporating any information received into its general body of knowledge. What was confidential is now suddenly in the public domain.
The Courts in common law jurisdictions in the United States, England and Australia apply the same principle that any public disclosure of otherwise confidential information or documents undermines any claim to confidentiality or legal privilege. In the United States v Heppner,[19] it was held the subject-matter of a party’s interaction with a public AI chatbot was not protected by legal privilege. The first English decision to confirm the effect of using AI on legal professional privilege came to the same conclusion about the loss of otherwise privileged material, once it was uploaded.[20] While not a case about AI, the nature of legal professional privilege was considered by the High Court of Australia in Glencore International AG v Commissioner of Taxation.[21] The applicant’s confidential documents were disclosed publicly without their knowledge or consent. The High Court found that the fact that legal professional privilege had attached to the documents before their disclosure, made no difference once the documents came into the public domain. This finding was so, despite the fact that Glencore itself had done nothing to waive their claim to legal professional privilege. There is little doubt that an Australian Court would uphold the Glencore decision in the event of a disclosure made to an AI platform.[22]
As soon as a client engages a lawyer, the first advice given to a client must be not to share any confidential or privileged information or documents with an AI platform.
4. What if a lawyer is unaware that AI is being used in preparing legal submissions?
The legal industry is rapidly developing AI tools to assist lawyers. This includes Closed AI tools, that is, an AI capability not open to the public and which segregate information from being shared publicly or from being used to train AI. Legal software providers promote their own bespoke AI platforms. Despite the comfort of a specialised product developed for the legal industry, one study by Stanford University found that hallucinations and other flaws still occur within specialist legal AI products.[23]
The presence or absence of a guarantee from a legal software provider will provide no solace if a practitioner relies on, without checking, a legal citation that turns out to be a hallucination. Murphy J in a native title dispute ordered the applicants’ solicitors to pay indemnity costs due to ‘… the firm’s use of artificial intelligence in the preparation of documents served on the respondents...’ [24] Murphy J made clear that a principal lawyer remains responsible for any incorrect citations, and cannot sheet blame, as in that case, to an inexperienced solicitor having carriage of a matter. His Honour cited with approval both the Dayal and Avianca decisions. The Legal Practitioners Liability Committee has issued important guidance to practitioners following Murphy J’s decision.[25]
The message is simple – responsibility begins and ends with the principal of a law firm for any improper use of AI generated research. This is so even if the principal has no personal involvement or knowledge about the improper use of AI. The point is that a principal lawyer must manage their practice by ensuring systems are in place to supervise the content of all written material submitted in the firm’s name.
D. A Hungarian Legal Perspective on the Regulation of mesterséges intelligencia (MI = AI)
The European Union has enacted comprehensive regulation of AI with the passage of the EU AI Act.[26] Hungary enacted domestic legislation, Act LXXV of 2025, in force since 1 December 2025 giving effect to the EU AI Act. The legislation classifies AI by four risk tiers, where AI is treated as either:
Prohibited;
High risk;
Limited; or
Minimal.
Comprehensive obligations are imposed on vendors of AI products to prohibit the exploitation of vulnerable users, to prevent scraping of facial images from the internet or CCTV footage, or otherwise not to operate in a manipulative manner. AI misuse has featured in Hungarian legal decisions. For example:
The Data Protection Authority[27] fined a bank €670,000 in 2022 due to its application of AI to assess the emotional state of customers calling the bank. That information was used to prioritise, which customer should first receive a call back. The data was held for 45 days.[28]
Like Company v Google Ireland C-250/25 concerns a copyright infringement case filed by a Hungarian media company against Google, with its European headquarters in Ireland. The allegation is that Google’s AI chatbot Gemini infringed the media company’s copyright by reproducing an article and further using the company’s content to train Google’s Large Language Model. The claim was filed at the Budapest Metropolitan Court.[29] The Court decided to stay the proceeding, and refer preliminary questions of European law to the Court of Justice of the European Union in Luxembourg.[30] Member States and the European Commission have intervened to make their own submissions. A decision is not expected until 2027.[31] The principles in dispute mirror the Australian debate on the difference between fair use to train LLMs desired by AI companies and what constitutes infringement of copyright matter owned by creators, such as authors, artists and musicians.
There are no reported examples of Hungarian lawyers misusing AI in legal proceedings before Hungarian Courts.
In March 2026, the Ministry of Justice launched www.jog.gov.hu, an official website that uses AI to answer the public’s questions about Hungarian law.
The Council of Bars and Law Societies of Europe has provided European wide guidance on AI use within the legal profession, that applies in Hungary.[32] The Hungarian Bar Association[33] has not yet developed a national code on the use of AI by its lawyers. On 25 September 2026, the First AI Summit organised by the European Judicial Training Network took place in Riga: ‘… to explore AI from legal, ethical, technological, and organisational perspectives in a multidisciplinary learning environment.’
No doubt, there will be a co-ordinated approach in due course to the development of legal regulation on the use of AI throughout the European Union, that will cover the Hungarian legal profession.
E. The use of AI in some Legal Practice areas in Australia
A lawyer needs to be aware of how a particular area of legal practice may have its own distinct set of issues arising from the advent of AI. These are just a few examples in some specialist legal areas.
Corporations Act
The Australian Institute of Company Directors summarised the shift in attitude among company boards to engaging with AI:[34]
As organisations globally embed artificial intelligence (AI) into their operations and business strategies, conversations in boardrooms are shifting from ‘how do we govern AI’ to ‘how can AI help us govern’?
The AICD noted certain concerns arising from AI use at a board level:
Box 1.2: AI for governance: red flags
1. Public AI tools, such as freely available web-based generative AI products, are used to analyse board packs or other sensitive information.
2. Directors have relied on an AI-generated summary of board papers without reading them.
3. Directors are using AI to perform the functions of management or in a way that undercuts management’s role (e.g. not raising questions of management and advisers which is part of directors’ duties of care and diligence).
4. AI tools are used as a substitute for effective board packs – surfacing fundamental information that should be provided by management.
5. AI outputs are not verified by a human and there is no ‘human in the loop’ – e.g. AI is used to generate minutes of a meeting without human review.
6. There are no controls on the level of organisational information directors can access using company-provided AI tools.
These concerns arising from the use of AI by directors undoubtedly undermine the value and purpose of a director sitting on a board. A director cannot consistently and independently act in accordance with their duties in equity and under the Corporations Act 2001 by delegating their independence to the random thoughts generated by an AI chatbot or other AI tool.
Intellectual Property Law
In patent law, it is a long established principle that an inventor must be named.[35] In Thaler v Commissioner of Patents[36] it was held at first instance for the first time in Australia that an AI intelligence system could be named as an inventor instead of a human creator.
This decision was unanimously overturned on appeal to the Full Court of the Federal Court, maintaining that an inventor must be a natural person to be named on a valid patent application.[37]
Employment law
AI has many applications in the employment relationship, including practices regarding recruitment, performance management, bonus entitlements, application of redundancy criteria, workplace surveillance, and decisions concerning dismissal. The biases inherent in AI can permeate all of these important workplace functions, if its human managers blindly apply the results of generative AI.
In the recruitment process for example, AI might be used to screen the suitability of job candidates. Bias unwittingly imported by the use of AI might not inform an employer who the right candidate for the job is. Instead, if AI relies on unlawful considerations such as gender, race, age, disability or other attributes, an employer might have contravened State of Commonwealth anti-discrimination law. In New York City, a law has been implemented to counter the use of AI in recruitment. Local law 144 requires an annual independent audit of automated hiring software to avoid the application of any unlawful discriminatory criteria by AI. The audit is intended to impose active human thought into the hiring process.[38]
AI tools can also be used to generate warning or dismissal letters, but AI cannot avoid liability for a range of employment related decisions under the Fair Work Act 2009 such as unlawful dismissal or adverse action. It is important that an employer using an AI platform ensures the tool is based on applicable and updated Australian law.
However, the climate towards AI use is not all negative. In one recent class action case brought under the Fair Work Act 2009, Lee J of the Federal Court urged the plaintiff’s lawyers to use AI to help calculate underpayments in the range of $100 million. His Honour made the following direction:[39]
7. By 6 November 2026, pursuant to s 37P(2) of the Federal Court of Australia Act 1976 (Cth), the parties are to:
(a) confer, with the assistance of a suitably qualified technology experts selected by the parties, as to the suitability of using artificial intelligence (AI) technology in the preparation of the matter for mediation and trial including, without limitation, the parties conferring regarding the use of AI technology in:
(i) collating and analysing quantification data for the purposes of mediation;
(ii) preparing chronologies and summaries of primary documents, including be reference to specified issues; and
(iii) managing and interpreting documents and other data that is relevant in the proceeding more generally,
(b) produce a report and/or draft protocol to the Court as to the results of that conferral;
(c) have reference, to the extent necessary, to the Federal Court of Australia, General Practice note - Use of Generative Artificial Intelligence Practice Note (GPN-AI), 16 April 2026, during that conferral.
His Honour went on to make some observations about how AI can be positively used by lawyers and judges alike:
2. To develop a related theme, a few days ago, I had the occasion of delivering the Sixteenth Sir Harry Gibbs Memorial Oration. In my speech All the Right Notes: Artificial Intelligence and the Future of the Common Law, I suggested that the real opportunity presented by artificial intelligence is considerably more interesting than whether a machine can draft a competent submission or find a case more quickly. It allows us to reconsider entrenched processes which developed in a world constrained by the capacity of human beings to collect, collate and synthesise information. Those constraints are disappearing rapidly and there is no reason to assume that procedures which developed in response to them possess some immutable virtue.
3. I also suggested that our lodestar should be the use of artificial intelligence to implement reforms which allow lawyers to spend more time exercising professional judgment and judges to spend more time identifying what really matters and deciding it according to law. That proposition has a very practical connexion with the overarching purpose. The just resolution of disputes according to law and as quickly, inexpensively and efficiently as possible is not furthered by requiring highly skilled solicitors, or armies of persons working under their supervision, to spend enormous amounts of time performing mechanical tasks of collection and synthesis if technology can perform those tasks reliably and at a fraction of the cost.
His Honour also referred to the importance of, and the ability of, AI to enable compliance with the overarching obligations placed on the lawyers to conduct litigation in an ethical, cost effective and time effective manner.
As for unrepresented litigants, there might be some limited cause for optimism in relation to a Fair Work Commission case this year. An unrepresented applicant successfully utilised AI to draft a winning submission that created a precedent on the question of a casual employee exercising a statutory right to convert to permanent employment.[40]
The unrepresented employee was a university academic teaching in the School of Computing Studies, and better skilled than many in engaging with the artificial intelligence.[41] What this case illustrates is that if AI is used competently, and perhaps with the benefit of some real technical skill is engaging AI resources, then unrepresented applicants might indeed have a free and viable alternative to engaging costly legal advice.
Privacy Law
The need for privacy law to catch up with technological change wrought by AI was best illustrated by a headline published on 11 September 2024 - Facebook admits to scraping every Australian adult user's public photos and posts to train AI, with no opt-out option. The admission was made by Meta during a Senate inquiry. The ABC noted that users in the United States and the European Union had the legal right to opt out from such activity, while there is no such right existed in Australia.
In 2024, the Privacy and Other Legislation Amendment Act 2024 (Cth) amended the Privacy Act 1988 to require reporting entities to include information relating to automated decision making in their privacy policies. From 10 December 2026, Australian Privacy Principle 1.7 will require an APP entity[42] to disclose in its privacy policy if it uses personal information as defined by the Privacy Act 1988 to make a decision that significantly affects the individual, and whether any personal information is used by any automated decision-making technology.
The privacy regulator, the Office of the Australian Information Commissioner, has cautioned businesses against disclosing personal information by use of AI, urging the highest standards of privacy governance.[43]
Work Health and Safety
New South Wales has become the first State to legislate for the use of AI at work in a manner that does not expose workers to the risk of injury.
The Work Health and Safety Amendment (Digital Work Systems) Act 2026 No 5 (NSW) amends the Work Health and Safety Act 2011 (NSW). It introduces a new definition of digital work system to mean an algorithm, artificial intelligence, automation or online platform and adds to a duty holder’s existing safety obligation not to expose the safety of a worker to a risk arising from a digital work system. The duty is expressed as:
A person conducting a business or undertaking must consider whether the allocation of work by or using a digital work system creates or results in any of the following risks—
(a) excessive or unreasonable workloads for workers at work in the business or undertaking,
(b) the use of excessive or unreasonable metrics to assess and track the performance of workers at work in the business or undertaking,
(c) excessive or unreasonable monitoring or surveillance of workers at work in the business or undertaking,
(d) unlawful discriminatory practices or decision-making in the conduct of the business or undertaking.
WorkSafe NSW has the full range of enforcement powers, including the right to demand an inspection of a digital work system. WorkSafe NSW is required to issue guidelines on workplace safety issues arising from AI in the workplace,[44] which are currently under development.
Property Law
A local cricket club in Queensland was concerned by the legality of a grant of a lease and the effect on its club members’ ability to use parkland to play games of cricket in the future. The club could not afford lawyers. Its president represented the club at Court, and used AI to draft legal submissions advancing the club’s interests.
The club failed at first instance. Costs were ordered against the club. Undeterred, the club’s president pressed on with an appeal. The Queensland Court of Appeal in a brief judgment dismissed the club’s application with costs, holding:
It may be acknowledged that the applicant’s written submissions to this Court were not characterized by some of the deficiencies which often bedevil submissions created by self-represented litigants through the use of Gen AI tools, in that they contained neither hallucinations nor confabulations nor misleading statements of fact. The problem for the applicant was that they failed to grapple with the critical point.[45]
The optimism of an unrepresented litigant achieving legal victory thanks to using generative AI, such as in the case of Baker v Macquarie University discussed above, is perhaps unfortunately the exception rather than the rule.
F. Regulation of AI Generally
The Commonwealth Government released its National AI Plan on 2 December 2025.[46] One initiative was the establishment of the Australian AI Safety Institute to monitor and advise on policy developments on the safe use of AI platforms.[47] Another key objective is to encourage AI-related investment into Australia.
Various areas of economic activity are being reviewed such as the effect of AI technology in relation to consumers. Similarly, the effect of AI on Australian workplaces has been reviewed. In March 2026, the House of Representatives Standing Committee on Employment, Education and Training conducted an inquiry into the effect of AI and technology on the Australia workplace: Australian Government response to the House of Representatives Standing Committee on Employment, Education and Training report: The Future of Work: Inquiry into the Digital Transformation of Workplaces. Among its conclusions, is that existing workplace laws already regulate the use of AI. The report found:
The Committee’s report raises a range of issues concerning the impact of AI in Australian workplaces. These issues are covered by existing workplace relations regulatory frameworks, including the Fair Work Act 2009 (Fair Work Act) and WHS laws. These laws are principle based and designed to be technologically neutral.[48]
The Government then established the Office of AI within the Department of the Prime Minister and Cabinet on 15 July 2026. Its aim is to: ‘… coordinate across Australian Government agencies to design and legislate the new Australia artificial intelligence (AI) standard and unlock AI training in Australia. These standards will include mandatory requirements for large AI data centres, including setting standards for energy and water, with strong copyright protections for Australian creators.’[49]
The most recent development is the issue of a Consultation Paper with regard to the development of AI data centres. The priority is for Australia to establishing a sovereign capability in all matters relating to AI:[50]
Australia has become one of the world’s leading destinations for data centre investment – but it must be on terms that work for Australians. Large data centres cannot drive up energy bills for Australian families and businesses, threaten scarce water resources, or be built without meaningful engagement with the communities in which they are located.
At the same time, we cannot afford to forgo data centre investment. It is essential to our national security, business productivity, science and innovation and resilience. With the right settings, and as a trusted Five Eyes member, Australia can be the infrastructure partner for our region.
AI is something much more than a new technological breakthrough. The Commonwealth government has cast AI as a matter of national survival, if Australia does not build its own AI capacity.
The sense of urgency is also reflected at a State level, with South Australia announcing its own Royal Commission into Artificial Intelligence in August 2026 with the purpose of ensuring: ‘…ensure South Australians maximise benefits and minimise unwanted impacts from the AI boom…’ [51]
Future regulation is inevitable given not only the technological and legal risks wrought by AI, but increasingly the emerging moral and ethical risks. Geoffrey Hinton, the Nobel Prize Laureate, often called the Godfather of AI, and responsible for key developments in the underlying technology of AI, has expressed repeated concerns about the safety risks posed by AI to humanity. Hinton is reported as saying:
"I think there’s a 50-50 chance it will get more intelligent than us in the next 20 years";
and
"We’ve never had to deal with things more intelligent than us. And so people should be very uncertain about what it will look like."[52]
Hinton backs the recent calls within the AI industry to slow down development of AI to ensure it is developed in a manner safe to humanity, before AI becomes more intelligent than humanity, and it is too late.[53]
Those working within the AI industry have also expressed concern, most recently with the resignation of Anthropic employee, Jacon Coxon, who is quoted as saying:[54]
" [AI corporations] are gambling with our lives. Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources.
and
"The people building AI earnestly believe that it could kill us all by the end of the decade. No other human activity poses this level of danger."
Important religious leaders too have expressed dire concern about the issues posed by AI, such as Pope Leo XIV.[55] In September 2024, the Council of Europe established a Framework Convention on Artificial Intelligence and human rights, democracy and the rule of law.[56] This is the first international treaty regulating AI.
The broader regulatory oversight and reform over AI will continue. The implication for lawyers is to maintain a current understanding of how their own practice might be impacted by the seeming unstoppable advance of AI.
G. What the Regulator says a Lawyer should do, and not do, with AI? What is the Future of AI in the Provision of Legal Services?
The Victorian Legal Services Board and Commissioner has provided guidance on how a lawyer should engage with AI:
Verify all AI-generated content against verified sources. If AI cites a case, it is your responsibility to verify the case exists and, importantly, that it is applicable to the argument or issue you are trying to run;
Avoid entering confidential client information into public AI tools. Privacy breaches may not only waive legal professional privilege, that can be disastrous for a matter, practitioners may face disciplinary action by doing so;
Follow Supreme Court of Victoria and County Court of Victoria guidelines on responsible AI use in litigation. These set out principles to be applied by litigations, including practitioners and parties, when using AI tools, which emphasis it is the duty of a party, and a lawyer, not to mislead the Court or another party.
However, a genuine pathway to make legitimate and beneficial use of AI in the legal sphere is emerging. The exhortations of Lee J in urging lawyers to make use of AI to make litigation more efficient is one example. Then, the success of an unrepresented academic in making an AI-generated submission to secure a legal victory on the regulation of casual employment shows consumers of legal services may be able to obtain effective legal advice through AI without ever seeing a lawyer. Though, the case of the cricket club blindly relying on AI to craft legal submissions does not bode well for unrepresented litigants unable to afford legal advice.
Lawyers must become technically savvy in order to use AI both ethically and effectively to discharge their basic obligations under the Legal Profession Uniform Law Australian Solicitors’ Conduct Rules 2015. An obligation imposed on lawyers to be IT literate is not a novel requirement. The Supreme Court of Victoria in its practice note, Technology in Civil Litigation almost ten years ago stated:[57]
The inability or reluctance of a lawyer to use common technologies should not occasion additional costs for other parties.
Sourcing technology services through a third party provider is accepted practice.
The use of common technologies is a core skill for lawyers and a basic component of all legal practice, whether provided in house or through a third party.
Those law firms which completely ban the use of AI may well deny their clients the legitimate benefits of AI, and deny their lawyers the opportunity to learn how to prepare for this next significant evolution in legal practice. What is clear is that AI is never a substitute for exercising legal due diligence at every stage of providing advice or representation.
The Board’s Statement on the use of artificial intelligence in Australian legal practice previously[58] informs lawyers of the application of long-standing ethical obligations as applied to the use of AI, which can be paraphrased as follows:
Maintaining client confidentiality. Lawyers cannot safely enter confidential, sensitive or privileged client information into public AI chatbots/copilots (like ChatGPT), or any other public tools. If lawyers use commercial AI tools with any client information, they need to carefully review contractual terms to ensure the information will be kept secure.
Providing independent advice. AI chatbots/ copilots and other LLM-based tools cannot reason, understand, or advise. Lawyers are responsible for exercising their own forensic judgement when advising clients, and cannot rely on the output of an AI tool as a substitute for their own assessment and analysis of a client’s needs and circumstances.
Being honest and delivering legal services competently and diligently. AI chatbots/copilots, research assistants, and summarisers cannot be relied on as a substitute for legal knowledge, experience or expertise. No tool based on current LLMs can be free of ‘hallucinations’ (i.e. responses which are fluent and convincing, but inaccurate), and lawyers using AI to prepare documents must be able and qualified to personally verify the information they contain, and must actually ensure that their contents are accurate, and not likely to mislead their client, the court, or another party.
Charging costs that are fair, reasonable and proportionate. Lawyers using AI to support their work should ensure that the time and work items they bill clients for accurately represent the legal work done by law practice staff for their client. Lawyers who use AI should ensure that it does not unnecessarily increase costs for their client above traditional methods (e.g. because of additional time spent verifying or correcting its output).
Whatever the future holds for AI, and that AI holds for humanity, AI has very much arrived for the legal profession. What is not new are the enduring ethical obligations to act in a client’s best interests and to act faithfully with regard to various overarching obligations owed to the court system.
Ignoring AI cannot be a strategy for any lawyer. Embracing education about the continuing traps posed by this emerging and at times sinister technology remains the only viable path to engage in effective legal practice.
What is inevitable, is that AI will remain a constant theme to be addressed, and enforced, by legal regulators. What is also inevitable is that lawyers will have to comply with increasing regulation, as the complexity of AI unfolds in the coming decades.
References
[1] Understanding the different types of artificial intelligence, https://www.ibm.com/think/topics/artificial-intelligence-types by the IBM Data and AI Team
[2] See China has cracked down on AI relationships. Is it ahead of the game? 1 October 2026, Stephanie Hegarty, Global population correspondent and Wanqing Zhang, Global China Unit, BBC, https://www.bbc.com/news/articles/cm4gjy9lr551o
[3] As the US Runs the AI Race, China Plays a Different Game, Meg Rithmire By Avery Forman, Harvard Business Review, 25 September 2026
[4] Frontier AI and Federal Cybersecurity Oversight: Where are we now on U.S. AI Regulation? by Natalie Donovan and Laura Houston Slaughter & May, and contributions from David J. Kappos and Ryan Wichtowski of Cravath, Swaine & Moore LLP, 1 October 2026
[5] H.R.9917 — 119th Congress (2025-2026)
[6] Other example of models include LCMs (Large Concept Models), VLMs (Vision-Language Models), SLMs (Small Language Models), MLMs (Masked Language Models), LAMs (Large Action Models), SAMs (Segment Anything Models)
[7] Research conducted by the Board into workplace culture in the legal sector in Victoria found 42% of respondents reported that long work hours had a negative impact on well-being. 881 lawyers completed the survey. See Understanding how workplace culture affects lawyer wellbeing, www.lsbc.vic.gov.au
[8] at [21]
[9] per Castel USDJ, Mata v Avianca, Inc, 448, Case 1:22-cv-01461-PKC Document 54 Filed 06/22/23, U.S. District Court for the Southern District of New York, cited by Humphreys J at [11]
[10] Dayal [2024] FedCFamC2F 1166 (27 August 2024)
[11] https://lsbc.vic.gov.au/news-updates/news/statement-mr-dayal-matter
[12] See Commissioner’s Message, which in turn refers to the VLSB’s Risk Outlook 2024 https://lsbc.vic.gov.au/lawyers/risk-outlook/risk-outlook-2024#2Artificialintelligenceimproperuse
[13] Source, Lawyer caught using AI-generated false citations in court case penalised in Australian first, The Guardian, 3 September 2025
[14] See https://www.damiencharlotin.com/hallucinations/
[15] 'AI slop' crowned word of the year 2025 in Macquarie Dictionary's committee and people's choice categories, Yiying Li, 25 November 2025, ABC News
[16] Ba v Sterling Parts Australia Pty Ltd [2026] FedCFamC2G 1245 at [1], 25 June 2026
[17] at [55]
[18] There are many articles on bias in AI, for example, 7 Common Types of AI Bias and How they Affect Different Industries,17 June 2025, www.lumenova.ai Bias in AI, Artificial Intelligence Hub, www.chapman.edu
[19] (S.D.N.Y. Feb. 17, 2026), per Rakoff DJ
[20] UK v Secretary of State for the Home Department [2026] UKUT 8, per Lindsley J
[21] [2019] HCA 26. Glencore’s privileged legal advice provided by its lawyers were leaked as part of the Paradise Papers affair, where investigative journalists published documents with legal advice. The ATO used the otherwise privileged documents against Glencore regarding its offshore movements of revenue.
[22] Your clients are using AI anyway, and it’s creating a privilege risk lawyers can’t control, Tyrilly Csillag, Law Society Journal, Online, 10 June 2026, Thomson Reuters
[23] Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools, Varun Magesh, Faiz Surani, Matthew Dahl, Mirac Suzgun, Christopher D. Manning, Daniel E. Ho, Journal of Empirical Legal Studies, Volume22, Issue2 at pages 216-242, June 2025
[24] Murray on behalf of the Wamba Wemba Native Title Claim Group v State of Victoria [2025] FCA 731
[25] Supervising AI, Legal Practitioners Liability Committee, 11 May 2026, https://lplc.com.au/resources/lij-article/supervising-ai
[26] Regulation 2024/1689 of the European Parliament and Council, 13 June 2024, governing the use of AI throughout the European Union. A Regulation under European law is directly applicable in each Member State without the need for local enabling legislation.
[27] Nemzeti Adatvédelmi és Információszabadság Hatóság
[28] Hungarian Data Protection Authority Issues Largest Fine To Date To A Bank For Unlawful Use Of AI, John O’Connor and David Cullen, William Fry LLP, 27 April 2022, Legal 500
[29] Budapest Környéki Törvényszék
[30] Az Európai Unió Bírósága
[31] CJEU to rule on AI and copyright in a landmark case against Google, Boriana Guimberteau and Pauline Desjours, Stephenson Harwood, 23 June 2025, https://www.stephensonharwood.com/insights/cjeu-to-rule-on-ai-and-copyright-in-a-landmark-case-against-google/
[32] For a fuller description of regulation of Hungarian lawyers in relation to AI use, see Artificial Intelligence 2026, Hungary Trends and Developments, E Szakacs, D Geyer-Hirt and J Talas, Danubia Patent & Law Office - Danubia Legal, Chambers & Partners, 21 May 2026, https://practiceguides.chambers.com/practice-guides/artificial-intelligence-2026/hungary/trends-and-developments/O26319
[33] Magyar Ügyvédi Kamara
[34] AI Use by directors and boards: Early Insights, AICD, December 2025
[35] Section 15 Patents Act 1990 (C’th)
[36] [2021] FCA 879
[37] Commissioner of Patents v Thaler [2022] FCAFC 62
[38] Local Law 144 of 2021, New York City
[39] Rogers v McDonald's Australia Ltd (AI-use) [2026] FCA 1264 (25 August 2026)
[40] section 66AAB Fair Work Act 2009
[41] Gregory Baker v Macquarie University [2026] FWC 3054, Commissioner Crawford
[42] Section 6(1) Privacy Act 1988
[43] GenAI tools in the workplace: balancing protection of personal information and business efficiency, Natalie Le, Director, 4 December 2025, Office of the Australian Information Commissioner
[44] See Schedule 1
[45] Beechmont Cricket Club Inc v Scenic Rim Regional Council [2026] QCA 187 (25 September 2026) at [3] per Bond JA.
[46] https://www.industry.gov.au/publications/national-ai-plan
[48] At page 14
[49] https://www.pmc.gov.au/domestic-policy/office-ai
[50] Getting it right: Building AI infrastructure that works for Australia: Consultation paper - September 2026, https://www.pmc.gov.au/resources/getting-it-right-building-ai-infrastructure-works-australia
[51] https://www.dpc.sa.gov.au/news/royal-commission-into-artificial-intelligence-announced
[52] '50-50 chance' that AI outsmarts humanity, Geoffrey Hinton says, Jon Erlichman, 14 June 2024, BBN Bloomberg https://www.bnnbloomberg.ca/business/economics/2024/06/14/50-50-chance-that-ai-outsmarts-humanity-geoffrey-hinton-says/
[53] 'Godfather of AI' Geoffrey Hinton backs Anthropic chief's call to slow down development, Oscar Coleman, 14 September 2026, ABC, https://www.abc.net.au/news/2026-09-14/godfather-of-ai-geoffrey-hinton-backs-ai-slow-down/107150010
[54] As AI behavior raises concerns, ex-researcher Jacob Coxon warns what may lie ahead, PBS Newshour, 17 September 2026, https://www.pbs.org/newshour/show/as-ai-behavior-raises-concerns-ex-researcher-jacob-coxon-warns-what-may-lie-ahead
[55] ENCYCLICAL LETTER MAGNIFICA HUMANITAS OF HIS HOLINESS POPE LEO XIV ON SAFEGUARDING THE HUMAN PERSON IN THE TIME OF ARTIFICIAL INTELLIGENCE, The Holy See, https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html
[56] https://rm.coe.int/1680afae3c
[57] Practice Note SC Gen 5 at [4.3], 30 January 2017
[58] https://www.lpbwa.org.au/artificial-intelligence-joint-statement
Disclaimer: The material on this website is for general information and educational purposes only. Its content and opinions do not constitute legal advice. Please consult a qualified legal professional if you require specific legal advice.
