The short answer is clear: Victorian law firms should not put confidential, sensitive or privileged client information into public AI tools such as ChatGPT.

That’s the practical message from the Legal Practitioners’ Liability Committee (LPLC), supported by guidance from the Victorian Legal Services Board and Commissioner, the Office of the Australian Information Commissioner (OAIC) and the Supreme Court of Victoria.

This doesn’t mean law firms must avoid artificial intelligence altogether. Generative AI can assist with appropriate, lower-risk tasks. However, your firm remains responsible for protecting client information, supervising staff and verifying every output.

If someone at your firm has already pasted client information into ChatGPT, don’t panic. Treat it as a potential information security incident, preserve the relevant records and obtain appropriate legal, privacy and cyber advice promptly.

Important: This article provides general information for Victorian law firms. It isn’t legal advice and shouldn’t replace advice from your professional advisers, insurer or regulator.

What does the LPLC say about confidential information and ChatGPT?

The LPLC advises practitioners to avoid entering client data or other sensitive information into AI tools.

Its updated guidance, “Limitations and risks of using AI in legal practice”, recommends limiting generative AI use to non-sensitive and publicly available information.

The concern isn’t limited to a client’s name. Confidential information may include:

• A client’s identity or contact details
• A unique factual scenario that could identify the client
• Draft advice, pleadings, contracts or correspondence
• Settlement positions and transaction details
• Commercially sensitive information
• Health, financial or family information
• Information obtained through discovery, subpoena or a confidentiality agreement
• Material subject to suppression or non-publication orders
• Legally privileged communications

A prompt can reveal a client even when you remove the client’s name. A combination of dates, locations, business details and unusual facts may be enough to identify the matter.

The LPLC also highlights the risk of breaching a non-disclosure agreement or client engagement term that restricts the use of AI or disclosure to third-party providers.

Is public ChatGPT different from a controlled business AI tool?

Yes, but a business subscription isn’t an automatic permission slip to upload client files.

Public AI tools are generally accessible over the internet. Their terms, settings, retention practices, integrations and data controls may not be appropriate for confidential legal work.

The Victorian Legal Services Board and Commissioner’s statement on AI in Australian legal practice says lawyers cannot safely enter confidential, sensitive or privileged client information into public AI chatbots or copilots.

The same statement says that if lawyers use commercial AI tools with client information, they need to carefully review the contractual terms to ensure the information will be kept secure.

That review should cover:

1. Who can access prompts, uploaded documents and outputs?
2. Will the vendor use customer data to train or improve a model?
3. How long is information retained?
4. Can the vendor or its subcontractors use the data for another purpose?
5. Where is the data stored and processed?
6. Can the firm control deletion and retention?
7. What happens if there’s a security incident?
8. Does the tool connect to web search, plugins or other third parties?
9. Can the firm audit user activity?
10. Does the vendor’s contract support the firm’s confidentiality and privacy obligations?

Some enterprise AI services state that business data isn’t used to train models by default. That can reduce one category of risk, but it doesn’t remove all risks. A firm still needs to assess the vendor, configure the service securely, control access and decide whether the proposed use is appropriate for the matter.

What privacy obligations should Victorian law firms consider?

Privacy compliance starts with understanding what information is being sent, where it goes and why it is being used.

The OAIC’s guidance on privacy and commercially available AI products says privacy obligations apply to personal information entered into an AI system and to AI-generated outputs containing personal information.

For firms covered by the Privacy Act 1988 (Cth), relevant Australian Privacy Principles may include:

• APP 3: Whether personal information is collected when using or generating information with AI
• APP 5: Whether individuals have been properly notified
• APP 6: Whether information is being used or disclosed for a secondary purpose
• APP 8: Whether information is disclosed overseas
• APP 10: Whether personal information is accurate
• APP 11: Whether reasonable security safeguards are in place

The OAIC recommends, as a matter of best practice, that organisations don’t enter personal information, particularly sensitive information, into publicly available generative AI tools.

That guidance doesn’t decide every question for every law firm. The relevant obligations will depend on the firm, the client, the information, the service and the circumstances. However, it reinforces why “everyone uses it” isn’t a sufficient risk assessment.

What does this mean for legal professional obligations?

Using AI doesn’t transfer professional responsibility to the software provider.

The Victorian Legal Services Board and Commissioner links AI use to existing obligations, including:

• Maintaining client confidentiality
• Providing independent advice
• Acting honestly
• Delivering legal services competently and diligently
• Supervising staff and legal work
• Charging fair and reasonable costs

The regulator’s statement explains that AI tools can’t replace a lawyer’s professional judgment. They can produce fluent but inaccurate content, and the lawyer must personally verify work before relying on it.

The Supreme Court of Victoria’s Practice Note SC GEN 25, issued in May 2026, provides further direction for court users.

It says:

• Information entered into a public AI tool could become publicly available.
• Confidential or sensitive information shouldn’t be entered into a public AI tool.
• Lawyers must maintain client confidentiality.
• Closed AI tools still require careful review of contractual terms and privacy settings.
• AI-assisted court content must be verified with meaningful human control.
• Lawyers should be able to identify AI-generated portions and explain how they were checked.

The position isn’t that every use of AI is prohibited. The position is that AI must operate inside a controlled legal workflow, with human judgment at the centre.

What should a law firm’s AI usage policy include?

A written policy turns informal experimentation into accountable practice.

The LPLC’s guidance on managing AI risks in law practices recommends clear risk management principles, policies and staff awareness.

At a minimum, your policy should explain:

Which AI tools are approved?

List the AI applications staff may use. Prohibit unapproved consumer tools for legal work involving confidential or personal information.

What information is prohibited?

Make the “do not upload” list practical and specific. Include client files, privileged material, identifying facts, health information, financial records, passwords and information covered by court orders.

Which tasks are permitted?

Lower-risk uses may include:

• Brainstorming a general structure
• Improving the tone of a generic email
• Creating a non-client-specific checklist
• Summarising publicly available material
• Generating fictional examples for training

Every output must still be checked. “Low risk” doesn’t mean “no risk”.

Who approves higher-risk use?

Require partner, risk or IT approval before staff use AI with firm or client information. Approval should be based on the specific tool, information and task.

How is AI use recorded?

Decide when staff must record the tool used, the purpose, the information entered, the output relied upon and the verification completed.

How will staff be trained?

A policy that nobody understands won’t protect the firm. Training should use realistic examples, including a prompt that appears anonymous but contains enough information to identify a client.

What should you do if confidential information has already been entered?

Respond quickly, but don’t make the situation worse by deleting evidence before assessing what happened.

Your response should generally include:

1. Stop further use of the relevant tool for confidential work.
2. Identify what was entered, when, by whom and from which device or account.
3. Record the exact prompts, documents and outputs, where possible.
4. Check the account type, settings, retention controls and vendor terms.
5. Restrict access and change credentials if there’s any concern about account compromise.
6. Notify the appropriate supervising partner or risk officer.
7. Contact your professional indemnity insurer or legal adviser for guidance.
8. Assess privacy, confidentiality, privilege, contractual and court-related issues.
9. Consider whether notification is required under applicable privacy, contractual or professional obligations.
10. Provide targeted staff training and update the policy after the review.

Don’t assume that deleting a conversation proves the information has been removed from every system or backup. Don’t assume that changing a setting retrospectively eliminates the original disclosure. Obtain advice based on the actual service and circumstances.

How can Victorian law firms use AI more safely?

The safest starting point is to separate innovation from sensitive client data.

Before introducing an AI tool, ask:

• Is this use necessary?
• Can the task be completed without client information?
• Can the information be genuinely de-identified?
• Is the tool approved by the firm?
• Have the vendor’s terms and data flows been reviewed?
• Is the information stored or processed overseas?
• Are access, retention and deletion controls available?
• Can a lawyer meaningfully verify the output?
• Does the use need to be disclosed to the client, court or another party?
• Has the firm trained and supervised the people using it?

A secure technology environment also matters. Strong identity controls, multifactor authentication, endpoint protection, data loss prevention, email security, backups and audit logging can help reduce the risk of accidental disclosure and unauthorised access.

Whole IT provides managed IT services for Melbourne businesses, including cloud management, cybersecurity and proactive support. We can help your firm review how staff access AI tools, where sensitive information is stored and which controls should be strengthened.

What is the practical takeaway for your firm?

Keep confidential client information out of public generative AI tools, and govern any approved business use with the same care you apply to other critical technology suppliers.

For Victorian law firms, responsible AI use means:

• Protecting confidentiality before chasing efficiency
• Reviewing vendor terms before uploading information
• Using firm-approved tools and settings
• Training lawyers and support staff
• Supervising AI-assisted work
• Independently verifying every important output
• Recording and disclosing AI use where appropriate
• Responding promptly to suspected data exposure

AI can help law firms work more efficiently. The opportunity is real. However, trust remains the foundation of legal practice.

If you’d like help reviewing your firm’s AI access, cloud environment, cybersecurity controls or data governance, contact Whole IT for a practical discussion about building a safer technology foundation.