Frequently Asked Questions

Can legal practitioners use artificial intelligence in legal practice?

Yes. Artificial intelligence, including generative AI, can be used in legal practice where its use is appropriate, responsible and consistent with a practitioner’s professional obligations. The South Australian Courts have recognised that generative AI may assist with efficiency, cost reduction and access to justice, but practitioners remain personally responsible for their work and conduct.

What is generative AI?

Generative AI refers to AI systems that can create content, such as text, images, audio, video or other material, in response to prompts. Common examples include tools that draft text, summarise documents, assist with research, translate material, analyse documents or generate images. The key issue for legal practitioners is not simply what the tool can produce, but whether its use is appropriate for the particular task and whether the output has been properly checked.

Can I rely on AI generated legal research?

AI generated legal research should not be relied on without independent verification. Generative AI can produce inaccurate, misleading or entirely fictitious authorities, citations, quotes or legal propositions. Practitioners should check any AI generated research against authoritative legal sources before using it in advice, correspondence, submissions or court documents.

Can I upload client documents into an AI tool?

Practitioners should be very cautious before uploading client information, confidential documents, privileged material, discovery, subpoenaed material or information subject to suppression, non publication or other restrictions. Unless the practitioner is satisfied about how the tool stores, uses, retains and protects the information, uploading that material may create confidentiality, privilege, privacy and court process risks.

The South Australian Courts’ guidance warns that unless a system is properly understood to be closed, uploading restricted material may involve a breach of obligations and possibly contempt.

Do I need to tell the Court if I have used AI?

It depends on the court, the type of use and any applicable practice note, guideline, rule or direction. In South Australian Courts, a party filing a legal document does not need to make a general disclosure merely because generative AI was used, but should be able to inform a judicial officer, if asked, about the use of AI.

In the Federal Court, disclosure is required in particular circumstances, including where AI has been used in a way that may affect evidence, admissibility or the Court’s use of the material.

Can AI be used to draft affidavits or witness statements?

Extreme care is required. A witness statement or affidavit must fairly and accurately reflect the witness’ own memory, knowledge and evidence. AI may assist with organising material, preparing chronologies or improving expression in appropriate cases, but it should not be used to alter, embellish, minimise or manufacture evidence.

Practitioners should not provide an AI generated draft to a witness in a way that risks suggesting the evidence the witness should give.

Can AI be used to summarise briefs, transcripts or large document sets?

Yes, but only with proper controls. AI can be useful for summarising large volumes of material, preparing chronologies, identifying topics in transcript or organising documents. Practitioners should check the output against the source material, ensure references are accurate and avoid uploading restricted or confidential material into tools that have not been assessed for confidentiality and security.

Can AI be used to alter, enhance or clarify evidence such as images, video or audio?

Practitioners should be very cautious. The undisclosed use of AI to alter, enhance or modify evidence before providing it to another party or presenting it to a court is inappropriate and may have serious consequences. Where AI is used to organise or improve the readability of material, its use should be transparent and the original material should be preserved.

Can an expert witness use AI?

An expert’s opinion must remain the expert’s own opinion, based on their specialised knowledge, training, study or experience. An opinion based on generative AI may be inadmissible if it does not expose the expert’s own reasoning. AI may assist with preliminary tasks such as sorting or interrogating data, but any use should be apparent from the report where relevant, and the expert or retaining party should be able to establish the reliability of the AI assisted work.

Should my practice have an AI policy?

Yes. A practice should consider adopting an internal AI policy that deals with approved tools, prohibited uses, confidentiality, privilege, privacy, verification of outputs, court disclosure, supervision of staff and record keeping. Practitioners should also review the terms of any AI product before use, including whether prompts or uploaded material are stored, reviewed by humans, used for model training, transferred overseas, retained after account closure or accessible by third parties.

What are appropriate lower risk uses of AI in a legal practice?

Lower risk uses may include administrative drafting, preparing first drafts of internal documents, generating checklists, organising non confidential information, improving plain English expression, or brainstorming issues for further legal analysis. Even for lower risk tasks, the output should be reviewed by a practitioner before it is used. The more a task involves legal judgement, client confidentiality, court material, evidence, personal information or professional responsibility, the more caution is required. 

Should a legal practice keep a record of when AI has been used?

Yes, where the use of AI is material to the work performed. A short file note may be appropriate recording the tool used, the purpose of use, the type of information entered, whether confidential or personal information was involved, and how the output was checked. This can assist with supervision, costs queries, client questions, court questions and later review if an issue arises. The Federal Court expects users to be able to inform the Court what generative AI was used, how it was used and for what purpose if required. 

Do I need client consent before using AI on a matter?

Not every use of AI will require express client consent, but practitioners should consider whether the proposed use affects confidentiality, privilege, privacy, costs, the nature of the work being performed or the client’s reasonable expectations. Express consent is more likely to be appropriate where client information will be entered into an AI system, where the tool is provided by a third party, where the use may affect costs, or where the client has raised concerns about AI use. 

Can AI be used to translate legal advice or client instructions?

AI translation tools should be used with care. Translation errors can change legal meaning, especially where instructions, advice, affidavits, settlement terms or court documents are involved. Practitioners should not assume that an AI generated translation is accurate, complete or culturally appropriate. Where accuracy matters, a suitably qualified interpreter or translator should be used, or the translation should be independently checked before it is relied on.

Can junior lawyers, graduates or support staff use AI tools?

Yes, but only within clear practice controls. A practice should specify which tools are approved, what information can and cannot be entered, what tasks are permitted, and who is responsible for reviewing outputs. The supervising practitioner remains responsible for ensuring that work is competent, accurate and consistent with professional obligations. Australian legal regulator guidance highlights the importance of risk based policies, supervision of junior and support staff, and review of documents containing AI generated content before they are finalised.

Can I charge clients for work assisted by AI?

Costs should accurately reflect the legal work performed. Practitioners should avoid charging as though a task was completed entirely by human effort if AI materially reduced the work required. At the same time, time spent properly checking, correcting and applying professional judgement to AI generated output may still be legitimate legal work.

What should I do if I discover that AI has produced an error after material has been sent or filed?

The practitioner should act promptly to assess the seriousness of the error, correct the record where necessary, and consider whether the client, another party or the Court needs to be notified. This is particularly important if the error concerns a legal authority, factual assertion, evidence, citation, quote, calculation or representation made to the Court. The Federal Court has stated that presenting false or inaccurate information to the Court is unacceptable and inconsistent with the responsibility not to mislead the Court or other parties.

Can a client waive legal professional privilege by uploading material to an AI tool?

Potentially, yes. Legal professional privilege depends on confidentiality. If a client uploads privileged material, such as legal advice, draft evidence, settlement strategy, counsel’s opinion, correspondence with their solicitor or documents prepared for litigation, into a public or third party AI tool, they may compromise the confidentiality needed to maintain privilege.

This does not mean every use of AI will automatically waive privilege. The risk will depend on the circumstances, including the tool used, the terms of use, who can access the uploaded material, whether the material is retained or used for training, and whether the disclosure is inconsistent with maintaining confidentiality. The difficulty is that once material is uploaded, it may be hard to know where it has gone, who can access it, or whether it can be deleted.

Practitioners should consider warning clients not to upload legal advice, court documents, draft evidence, discovery, confidential correspondence or other sensitive material into AI tools without first obtaining legal advice. This warning may be appropriate in retainers, client updates, litigation guidance, or matter specific correspondence. Practitioners should also consider giving similar guidance where clients are self managing parts of a matter or using AI to understand advice, prepare documents or summarise legal material. Current commentary on AI and privilege emphasises that confidentiality is central to legal professional privilege, and that using third party AI tools can create real risks for privileged and confidential material.

Back to AI Hub