Beware ChatGPT – It Is Not Your Lawyer
There has been an increase of enquiries to Potts Lawyers from people who have either consulted Artificial Intelligence (AI) or ChatGPT about their legal issue, or have utilised AI to prepare a claim or demand, or respond to an issue without success.
We get it – knowledge is power, and lawyers are expensive.
The problem we are finding, however, is that ChatGPT has significantly harmed the enquirer’s position and we are then engaged to wind back that harm, which causes us to work on your matter for a lot longer than would have been necessary, and ultimately costs you more money.
Artificial Intelligence Hallucinates
This is an evolving area, however it has become clear in recent times that AI cannot be trusted to give you sound legal advice. ChatGPT hallucinates.
The Queensland Civil and Administrative Tribunal has published guidelines that relevantly warn:
Anyone using AI must bear in mind that Generative AI chatbots cannot distinguish between facts, inferences and opinions contained in their source datasets. This means that text which they generate in response to a prompt may contain incorrect, opinionated, misleading or biased statements presented as fact.
These programs, especially the most popular ChatGPT, pretends to find correct legal answers and sometimes, try and please the user by creating false authorities or case law to support up its answers. This is popularly referred to as AI hallucinations but is known simply to the Court as false and misleading submissions.
A known reliability issue is that while many AI programs can ‘think deeply’ and eventually arrive at an acceptable position or acknowledge an earlier error, the energy and computing resources required to do this mean that they are often set up to provide you with a superficial answer at first and only when prompted numerous times will they agree that their superficial answer is incomplete or misleading. Even then, they may confidently give you the wrong answer with a view to please the user, rather than provide accuracy.
Even if you tell ChatGPT that it has given you a false answer, it may say ‘You’re right, thanks for calling that out’ but you first need to know the law to be able to distinguish the inaccuracies. If you directly ask ChatGPT whether it can be relied upon for drafting legal submissions in Queensland, it will concede that it cannot be relied upon.
No Confidentiality
Of course, all the client work we undertake is confidential and legally privileged to some extent.
However, if you upload documents (such as draft submissions or draft correspondence to ChatGPT or another AI program) this confidentiality cannot be assured. Even ChatGPT will tell you that it is a bad idea to upload confidential or privileged material to ChatGPT and warns you that you may be breaching your privacy or your organisation’s data security policy before uploading. It specifically warns lawyers to never upload any documents that include ‘any possible personal identifiers’.
As such, lawyers cannot ethically use AI or ChatGPT to review client related documents. Any such use is a potential breach of their duties of confidentiality and legal professional privilege.
We warn however, that you should not use AI to review your legal documents and issues either.
By uploading material, you are essentially sharing it to a third party waiving confidentiality and legal privilege, as ChatGPT will store the information. Even if you delete the file from the Chat, it warns that it may retain a backup for ‘safety and systems integrity’ purposes.
Lawyers Caught Out
The problem is so widespread that lawyers who have been utilising AI to assist their practice have been caught out providing the wrong advice, with made up quotes from concocted case law and misinterpreted legislation.
This has caused the courts and tribunals to issue a practice direction so that all lawyers who use AI in preparation of their oral or written submissions must:
- verify themselves the accuracy and relevancy of references to legislation, authorities and sources; and
- ensure that their document and submissions are expressed in a way that reflects their ethical obligations to the Court.
Any error with respect to using AI assisted written or oral submissions that uses incorrect or inaccurate references that wastes the Court’s time or harms confidence in the justice system may result in the referral of the lawyer responsible to the Legal Services Commission for prosecution.
Where To Now?
As at the date of publication of this article, the better view in the AI industry is that AI will never be able to get rid of fake results or hallucinations.
If AI was to be programmed to strictly say ‘I don’t know’ instead of guessing, it is estimated that 30% of queries would end up with the answer ‘I don’t know’. AI companies say that this would ruin the user experience so much that their companies (with billions in capital) would become worthless. So, for the foreseeable future, you can expect to contend with fake results to your legal queries.
AI is good (but arguably not that much better than Google) at giving very broad tips about the legal process. For example if you ask ChatGPT for tips because you are feeling nervous about an upcoming Court matter it will give you very sensible tips about how to prepare for the attendance and managing nerves on the day by breathing techniques and shifting your mindset.
However, the current state of AI in providing reliable legal answers is such that we cannot use AI to replace the hard work of marrying up your individual factual, financial and psychological situations with the law – let alone taking into account known preferences of the decision maker we are appearing before in Court, which might require a careful balance of where the relevance and emphasis is placed on any given argument we are attempting to develop.
You should not rely on AI for your legal problem either.
Take your lead from the concerns set out in the Courts’ practice directions, which focus on a lack of accuracy and reliability in current AI offerings. Such AI induced errors, even when accidental, are not in keeping with your obligations to the Court to be honest, truthful and assist the administration of justice. Moreover, AI hallucinations harm your legal position and can potentially have serious implications if you do not obtain the correct advice.
It is always best to contact a lawyer who has the necessary legal training and experience to be able to consider your legal issue and the options available to you to adequately protect your interests. While we understand legal fees are expensive, the cost of rectifying the AI hallucinations and its implications on your legal position can far exceed the cost of engaging a lawyer in the first instance.
Ensure your confidentiality is maintained and legal advice is accurate by calling Potts Lawyers today.