Legal Practice Director Jeff Garrett sets out two of the biggest risks that clients face when they use AI on their own: compromised confidentiality, and higher legal costs that don’t add any real value to the case.
I read a piece in Crikey recently by Michael Bradley, describing how the legal profession is “drowning in the tide of AI,” and how the courts are becoming inundated and strained under its weight. It certainly resonated with me and is an issue we are seeing more often in day-to-day legal practice.
Across all of our departments, but particularly in our litigation, family law and Wills & estates practice areas, we are seeing two issues repeatedly show up: the risks to confidentiality and the unnecessary costs clients can rack up by running questions through an AI tool like ChatGPT or Claude and then bringing those results to us to validate or ‘revise’.
This year marks 45 years since I started at Attwood Marshall Lawyers in 1981, and in that time, I have watched and experienced many new technologies change the way we work. Most of them made our work faster and more accurate. AI is different because it is changing how clients think about their own legal matters, and not always for the better.
When you start seeing clients place as much trust in an output from ChatGPT as they do in the legal advice they receive (or more!), that is concerning. A cautionary note to all – AI is still developing and makes a lot of mistakes. You have to really know your stuff to pick the incorrect parts of the AI generated material.
One: The confidentiality problem
When a client takes legal advice we have given them (whether that is a letter, a strategy note or a draft document) and feeds it into a public AI tool to have it summarised, questioned, or reworked, that information is being disclosed to and processed by a ‘public’ third-party system outside the lawyer’s control.
Depending on the platform, its settings and terms of use, that information may be retained or used in ways the client does not fully appreciate. A client may think they are simply asking a private question, without considering where that information goes, how long it is kept, or how it may be used. You could find, to your chagrin, that the facts of your case and the confidential legal advice you have received have been retained by the provider or, depending on the settings, used to train future versions of the tool. That material could also be exposed through shared links, data breaches or court orders requiring the provider to preserve user conversations.
There is also the issue of legal professional privilege. Voluntarily disclosing privileged legal advice to a third-party AI system can, in some circumstances, put that solicitor-client privilege at risk. Whether uploading advice to an AI platform would amount to a waiver would depend on the unique circumstances of each case, and I would be reluctant to predict how a court would ultimately determine that question.
However, it creates an unnecessary risk, and it is not one I would want a client to discover the hard way in a hard-fought litigious matter. It could decide a case or compromise your positioning and leverage to your detriment.
Two: The cost problem
When a client uses AI tools to generate a long list of questions, propose extensive amendments to a draft, or query every element of the legal advice given to them in a consultation, a lawyer still has to read, assess and respond to all of it.
Most lawyers work on an hourly rate, so if they are presented with more questions and further drafts, it means more billable time, whether or not the AI-generated material actually adds anything useful to the matter.
We understand the appeal. The content that these AI tools produce looks thorough and convincing on the surface. But most of it does not hold up once you dig deeper.
Clients often think they are fast-tracking their matter by using AI to draft documents or answer initial questions. But the output is only as good as the information and context put into it, and that is where problems arise.
I encourage you to bring your genuine questions to your lawyer, instead of asking us to sort through a document that AI produced. It will save you money and get you a better, more accurate result.
The rise of the self-represented litigant
Beyond the change in lawyer-client relationships, there is a bigger shift happening across the justice system. Self-represented litigants have always existed, but AI has considerably changed the scale.
The Fair Work Commission recently reported a proliferation of self-represented litigants using AI tools in unfair dismissal and general protections matters, and the courts are more broadly seeing a wave of AI-drafted claims and submissions.
This year, the Federal Court released guidelines on the ramifications of citing imaginary cases or legislation in court documents.
Courts afford self-represented litigants considerable latitude, granting adjournments and explaining procedural requirements. The represented party’s lawyers can spend significant time on correspondence and hearings that would have been straightforward between experienced practitioners.
Often, they fail. A litigant using AI to prepare their case does not know whether the output is legally sound, and the AI has no way of knowing either. What looks like a well-prepared case is often built on shaky foundations, which tends to cost the litigant more in the end, not less.
We have noticed a marked increase in self-represented litigants generally, which usually makes it very difficult to efficiently deal with contested matters.
Our advice to clients
AI has a role in many industries, including legal practice. Our teams use tools designed specifically for lawyers for certain tasks and work hard to check against source material, correcting anything that does not measure up. That process works because a lawyer with genuine subject-matter expertise is directing the tool and checking its output at every stage.
A client or self-represented litigant using AI to generate legal strategy or court documents without that oversight is in a very different position. They are relying on a chatbot’s best guess of their unique situation.
These AI tools can provide great summaries of legal issues in general terms.
But when it comes to your own matter, bring your questions directly to your lawyer rather than routing them through an AI tool first. It protects your confidentiality, it keeps your costs down, and it means the advice you get is based on your actual circumstances, up-to-date legal knowledge and state-specific legislation.
The fundamentals of good legal advice have not changed: experience, confidentiality, professional judgment, a genuine understanding of your matter, and the care and consideration that comes from having a real person know what is at stake.
As the Cruel Sea song says: Better get a lawyer, son. You better get a real good one!
