Blunders of using ChatGPT in Legal Research

In the news recently was a case of a lawyer that used ChatGPT to prepare for a trial only to find out that chatGPT had invented a set of cases that were cited by the lawyer. In the past few months of extensive use of chatGPT it appears that the ‘hallucinations’ of the AI model are in fact quite common and appear in many conditions and cases without the system displaying any caution or warnings on these hallucinations.

This behaviour is actually normal. ChatGPT as we know is a ‘large language model’ and the logic behind it focuses more on predictive expression and not… truth. So, using this approach as part of a trial really is a dangerous way of handling the case as – obviously – it can lead to disasters.

This case has been used repeatedly as an example that shows that AI is simply not ‘there’ for using it as part of legal technology and although this is partly right, it is not exactly right. AI in legal technology can have many applications already. Large language models even can provide the substructure for legal cases that will be edited modified and updated to apply to the specific case that is at hand.

AI works in legal technology hand in hand with the natural intelligence and experience of the lawyer and at this stage at least it does not pose a threat of replacing the lawyer – no matter what the hype is.

This collaboration between AI and the lawyer we believe is part of the new way of working for the new lawyer. This is exactly how Lawptimize operates as well and aims to combine the computational and predictive power of machine learning, Bayesian analysis and game theory with the knowledge and experience of a lawyer in order to produce the best and most effective results for each case that is being analyzed by the platform.

Litigation Risk Management

Request a Demo