It has become routine for an AI note-taker to join a Teams or Zoom call uninvited, and for an email offering everyone on the call access to the generated summary to land in your inbox soon after. Convenient, certainly. But if legal matters are discussed while that bot is listening, the convenience can come at a real cost.
Legal professional privilege only protects communications that are confidential. Once a business understands how AI note-takers and chat tools actually handle the information fed into them, it becomes clear why this convenience deserves a second look.
What Is Legal Professional Privilege?
Legal privilege exists to let clients speak openly with their lawyers, without fear that those conversations will later be forced into the open — for example, produced to a court under a discovery order. It generally covers two situations:
- Legal advice privilege — protects confidential communications made for the dominant purpose of getting or giving legal advice
- Litigation privilege — protects confidential communications and documents prepared for the dominant purpose of actual, anticipated, or pending legal proceedings
Confidentiality sits at the centre of both. If a client behaves in a way that is inconsistent with keeping a communication confidential, the privilege can be lost — regardless of whether the client meant to give it up.
Does an AI Platform Keep Your Information Confidential?
Publicly available AI tools — the well-known large language models included — commonly retain what a user submits and can use that material to further train the underlying model. Every meeting bot, chat assistant, and drafting tool built on top of these models inherits the same underlying question.
These tools are genuinely useful: agendas written up automatically, dense documents condensed, emails tidied into something more polished. What is less obvious to most users is exactly how each platform treats the data once it is submitted.
For anything that might otherwise be privileged, the question becomes unavoidable: is information still confidential once it has passed through an AI system you do not control?
A US Case Worth Watching: United States v Heppner
No Australian court appears to have ruled on this point yet. A recent US decision, however, offers a useful preview of how a court might approach it.
In that matter, a defendant used a public AI chatbot on his own initiative, without instruction from his legal team, to generate material analysing his potential defence. He later shared what the tool produced with his lawyers and then sought to claim privilege over it. The court refused.
The reasoning is instructive. The court found, broadly, that:
- No lawyer-client relationship exists between a person and an AI platform
- Because the AI tool was publicly accessible and had no arrangement in place to keep submitted data private, communications with it could not be treated as confidential
- The AI-generated material did not meet the dominant purpose test, since it was not prepared at the direction of the lawyers
- Feeding material into a generative AI tool was, in effect, disclosing it to a third party
- Even though the original information was likely privileged before it was uploaded, submitting it to a publicly accessible platform amounted to a waiver
The facts concerned a chatbot rather than a meeting bot, but the court's logic applies just as readily to transcription services and other AI tools that process business conversations.
What This Means for Meeting Bots and Transcription Tools
Meeting bots and transcription services work the same way underneath: an audio recording is uploaded to an AI platform, which then generates the transcript, summary, action list, or follow-up email using a large language model.
Not every one of these services is built with confidentiality and data security as a priority, or has the safeguards needed to keep data private and privileged. If legal advice or matters connected to a current or anticipated dispute come up during a recorded meeting, a party to future litigation — or a regulator — may seek access to that transcript. Depending on how that particular platform handles data, privilege may not be available to stop them.
Practical Steps for Businesses
A few sensible habits go a long way toward keeping confidential and privileged material protected:
- Check the privacy settings of any AI tool you use. Favour platforms with clear, contractual commitments that your data will not train their models, will not be shared with third parties, and will be stored securely.
- Read the terms of any AI transcription service before relying on it. Understand exactly how it uses and stores what you feed into it.
- Prefer enterprise-grade AI platforms with strong data protections over free, publicly accessible tools when confidentiality matters.
- Put an internal AI-use policy in place so staff know which tools are approved, and make clear that legal advice should never be disclosed to platforms outside that list.
- Turn AI note-takers off when legal advice is being discussed — whether that is a meeting with your lawyers or an internal discussion among staff.
- Ask your lawyer before using AI in connection with a legal issue. This applies to disputes, investigations, regulatory matters, or anything else that is meant to stay confidential.
- Train your team on what legal privilege is — how it can be waived, and how privileged material should and should not be shared internally.
Staying alert to these habits is the difference between privilege that holds and privilege that is lost without anyone meaning to give it away.
How Phan Campbell & Associates Can Help
At Phan Campbell & Associates in Footscray, we advise Melbourne and Victorian businesses on protecting privileged and confidential information as workplace technology evolves — including how AI tools fit safely into everyday operations.
If your business is unsure how AI use might affect legal privilege or confidentiality, we are happy to talk it through.
Frequently Asked Questions
1. Can using an AI meeting bot waive legal privilege?
Potentially, yes. Privilege depends on a communication staying confidential. If a meeting bot processes a conversation through a public AI model without adequate safeguards, that can count as disclosure to a third party — risking a waiver of privilege whether or not it was intended.
2. Has an Australian court ruled on AI and legal privilege?
Not yet, as far as we are aware. The clearest guidance so far comes from a US decision, Heppner, where a court declined to recognise privilege over material generated using a public AI chatbot without the defendant's lawyers involved. Its reasoning is a useful indicator of how Australian courts might approach similar facts.
3. Are all AI tools equally risky for confidential and privileged information?
No. Risk largely depends on how a platform handles data — whether it trains its model on your input, whether it shares data with third parties, and what security safeguards apply. Enterprise-grade tools with clear privacy terms carry materially less risk than free, publicly accessible AI tools.
4. What should a business do before using AI around legal matters?
Check the privacy terms of any AI or transcription tool, avoid running it during discussions involving legal advice or disputes, put a clear internal AI-use policy in place, train staff on what privilege means and how it can be lost, and ask your lawyer before using AI in connection with a legal issue, investigation, or dispute.