Artificial intelligence tools are increasingly being used in legal practice to assist with research, drafting, summarising, document review, transcription, practice management and other legal work.
While AI may assist legal practitioners, it does not replace professional judgment, ethical responsibility or the duties owed to clients, courts and the administration of justice. Practitioners remain responsible for the work they produce, the advice they give and the way client information is handled, whether or not AI has been used.
This page brings together key issues practitioners should consider before using AI in legal practice, including competence, verification, confidentiality, privilege, transparency, billing, litigation use, supervision, record keeping and practice governance.
AI does not change your professional obligations
The use of AI does not alter a practitioner’s professional obligations under the South Australian Legal Practitioners Conduct Rules.
Practitioners remain responsible for exercising independent professional judgment and for ensuring that any AI assisted work is consistent with their duties to clients, the court and the administration of justice.
Relevant conduct rules include:
- Rule 3, paramount duty to the court and the administration of justice
-
Rule 4, fundamental ethical duties
-
Rule 8, client instructions
-
Rule 17, independence and avoidance of conflicts with the administration of justice
-
Rule 19, candour to the court
-
Rule 22, communications with opponents
Practitioners should ensure AI use does not compromise professional independence, result in misleading conduct, or diminish the practitioner’s responsibility for the work produced.
Before using an AI tool in legal practice, practitioners should have a reasonable understanding of what the tool does, what it is suitable for and what its limitations are.
Generative AI tools can produce inaccurate, incomplete or fabricated information. They may also produce outputs that appear confident or authoritative, even where the content is wrong.
Practitioners should consider:
-
what the tool is being used for;
-
whether the tool is suitable for that task;
-
whether the tool relies on current and authoritative information;
-
whether the output can be independently checked;
-
what data is being entered into the tool;
-
whether client confidentiality and privilege are protected.
Check and verify AI outputs
AI outputs should not be relied on without independent review and verification.
This is particularly important where AI is used for legal research, drafting advice, preparing submissions, summarising evidence, identifying legal authorities or producing correspondence.
Practitioners should check AI outputs against authoritative sources, including:
-
current legislation;
-
reported decisions and authorised databases;
-
court rules, practice notes, practice directions and guidelines;
-
client instructions and source documents;
-
the practitioner’s own professional judgment.
Protect confidentiality, privacy and privilege
Practitioners should be cautious before entering client information into any AI tool.
Client information may include confidential information, personal information, privileged material, commercially sensitive information or information relating to a client’s matter.
Open or public AI tools should not be treated as secure merely because they are widely available or convenient to use. Before entering client information, practitioners should consider whether the tool has appropriate contractual, technical and organisational controls to protect the information.
Issues to consider include:
-
whether user inputs are stored, reviewed or used to train the system;
-
whether the tool is public, enterprise, private or locally hosted;
-
whether access controls and multi factor authentication are available;
-
whether data is encrypted in transit and at rest;
-
whether the provider gives clear information about data handling and retention;
-
whether client consent is required or appropriate;
-
whether using the tool could risk waiver of privilege;
- whether there are any other legal or procedural restrictions on the use or disclosure of the information, including obligations arising from court rules, confidentiality orders, implied undertakings or the Harman principle, which may restrict the use of documents or information obtained through compulsory court processes for purposes outside the proceeding.
Relevant conduct rules include Rule 9, which concerns confidential client information, and Rule 10, which concerns duties relating to former client confidential information.
Where possible, practitioners should adopt data minimisation practices and avoid entering unnecessary identifying, sensitive or privileged information.
Be transparent and accurate
Practitioners should not misrepresent the role AI has played in legal work.
Depending on the circumstances, transparency may be relevant to communications with clients, courts, other practitioners or third parties. Practitioners should also be aware that some courts and tribunals may have specific requirements or expectations about AI use, including disclosure requirements.
Practitioners should consider:
-
whether AI use needs to be disclosed;
-
whether a court, tribunal or practice direction applies;
-
whether the client should be informed about the use of AI;
-
whether AI use affects billing or costs disclosure;
-
whether any AI assisted work has been sufficiently reviewed before being used.
Relevant conduct rules include Rule 19, which concerns candour to the court, Rule 22, which concerns communications with opponents, and Rule 23, which concerns dealings with unrepresented persons.
Billing and costs
The use of AI does not, of itself, justify charging a client as though work was performed entirely without technological assistance.
Costs should remain fair, reasonable and proportionate. Practitioners should ensure billing practices accurately reflect the work actually undertaken, including the professional review, analysis, supervision and judgment applied to AI assisted work.
As AI tools become more capable and widely adopted, practitioners may also need to consider whether there are circumstances where the appropriate use of AI or other technology may assist in delivering legal services more efficiently and at lower cost to the client. This does not mean that AI must be used in every matter, or that efficiency should come at the expense of professional judgment, confidentiality, accuracy or proper supervision. However, where a task can be performed safely and appropriately with technological assistance, practitioners should be alert to whether traditional methods remain reasonable and proportionate in the circumstances.
Where AI related costs are charged as disbursements or expenses, they should be disclosed clearly and charged only where permitted under the costs agreement and applicable law.
Relevant conduct rules include Rule 4, including the duties to act honestly and deliver legal services competently, diligently and promptly. Practitioners should also consider their costs disclosure and billing obligations under the Legal Practitioners Act 1981 (SA).
AI in litigation and court work
AI may assist with litigation tasks such as drafting, summarising, document review, legal research and case preparation. However, practitioners must ensure that AI use does not compromise duties owed to the court or the integrity of court processes.
Particular care should be taken with:
-
legal authorities;
-
citations;
-
quotations;
-
summaries of evidence;
-
affidavits and witness statements;
-
expert material;
-
submissions and court documents.
Relevant conduct rules include Rule 3, Rule 19 and Rule 20.
Practitioners should also check whether the relevant court or tribunal has issued any AI specific guidance, practice note or direction. This is important because requirements may differ between jurisdictions and forums. Some courts or tribunals may impose specific obligations when AI is used, including restrictions on the use of AI for particular materials, requirements to verify AI generated content, requirements to disclose the use of AI, or requirements to certify that material has been independently checked by a legal practitioner. Practitioners should ensure they understand and comply with any applicable requirements before using AI in litigation or court related work.
AI in advisory, transactional and other legal work
AI use is not limited to litigation. It may also arise in commercial, regulatory, transactional, succession, property, family, employment and other legal work.
Practitioners should ensure that AI assisted work is accurate, tailored to the client’s circumstances and independently assessed before being provided to a client or relied upon.
AI should not be used to provide unsupervised or automated legal advice directly to clients. Practitioners should also consider whether the use of AI in this way is covered by their professional indemnity insurance. Depending on the structure of the service, automated or unsupervised advice may fall outside the scope of cover, particularly where advice is generated or delivered without appropriate practitioner review, supervision and professional judgment.
Relevant conduct rules include Rule 4, Rule 8, Rule 9 and Rule 17.
Supervision, training and practice governance
Legal practices should consider adopting internal policies or procedures for AI use.
The level of governance required will depend on the size of the practice, the types of AI tools being used, the sensitivity of the information involved and the nature of the legal work being performed.
Practices should consider:
-
who is permitted to use AI tools;
-
what tools are approved for use;
-
what information must not be entered into AI tools;
-
when client consent may be required;
-
how outputs are to be checked;
-
how AI use should be recorded on the file;
-
how junior lawyers and staff are supervised;
-
how tools are reviewed over time.
Relevant conduct rules include Rule 37, which concerns supervision of legal services, and Rule 4.1.3, which concerns the competent, diligent and prompt delivery of legal services.
Record keeping
Where AI is used in client work, practitioners should consider whether a file note noting the use of AI is appropriate.
The level of detail should be proportionate to the significance of the AI use. A brief record may be sufficient where AI is used for low risk drafting or administrative assistance. More detailed records may be appropriate where AI materially contributes to legal analysis, advice, evidence, court documents or other substantive work.
A file note may record:
-
the AI tool used (including the 'model' used);
-
the purpose for which it was used;
-
whether client information was entered;
-
whether information was anonymised or limited;
-
how the output was checked;
-
what changes were made before the work was relied upon.
Maintaining appropriate records can assist practitioners to demonstrate that AI assisted work was reviewed, verified and subject to professional judgment.
Before using AI in a legal matter, ask:
-
Is this an appropriate task for AI?
-
Am I using a secure and suitable tool?
-
Am I protecting client confidentiality and privilege?
-
Have I checked the output against authoritative sources?
-
Have I applied my own professional judgment?
-
Do I need to disclose the use of AI?
-
Should I record the use of AI on the file?