AI in Legal Proceedings: Efficiency Booster or Risk Factor?

Once a feature of fiction, Artificial Intelligence (AI) has now become a much more permanent fixture in our everyday lives, which is evident in that it now potentially provides a game-changing tool for realising efficiencies in construction, infrastructure and energy disputes which are data and document-heavy. The question is whether our court and tribunal procedures and rules are ready for the advent of AI.

The Technology and Construction Court (TCC) has always been a forward-looking court and in 2013, it implemented a protocol for e-disclosure as part of its Court Users Guide. That protocol endorsed the use of predictive coding, an early frontrunner of modern AI tools, to reduce the cost and burden of the disclosure exercise by combining human decision-making with iterative machine learning.

Following this lead, the Chancery Division of the High Court ordered in David Brown v BCA Trading Limited & Ors [2016], the use of predictive coding for disclosure instead of traditional keyword searches. More recently, following the roll-out of the Disclosure Pilot Scheme, the Civil Procedure Rules adopted predictive coding as part of PD57AD.

The use of AI in this way is not confined to the courts of England and Wales however, according to a recent arbitration survey by colleagues at Bryan Cave Berwin Leighton Paisner LLP, 30% of respondents had used or permitted the use of AI in document review and production. The trend towards limiting or controlling document production in some way (such as specific documents or categories of documents) means AI will have an increasing role.

In both forums, there is, or will be, an emphasis on the parties collaborating or reaching an agreement on matters such as which software, analytical techniques and coding strategies should be used. Under the Civil Procedure Rules, the court may give directions to this effect, and it is likely that tribunals would do the same in arbitration if the parties agreed to use AI. This results in a transparent and accountable process, which is important because it provides certainty. It is also where the Tribunal’s own knowledge and willingness to adopt technology will be important because as with all disputes, robust case management is key to efficient resolution.

But what happens in arbitration when there is no agreement or order from the Tribunal, and yet one party has still utilised  AI as part of document production? If discovered later, the other party may have legitimate concerns about the reliability and accuracy of the AI deployed, raising questions about the enforceability of an award. While traditional document production may face similar criticisms regarding reliability and accuracy, it might be easier to identify and quantify errors resulting from the use of AI, thus undermining the process. Already, the media is reporting on instances of large language learning models being inappropriately deployed in court proceedings, generating fictitious authorities and opinions.

Time will tell whether such concerns are realised, but with all matters, the lesson is one of appropriate deployment with appropriate supervisions, checks and balances. We expect the appropriateness and use of AI to be tested in courts in due course providing valuable guidance to users. That is before we even begin to consider the potential use of AI as tribunals for certain types of disputes!

Posted by: Oliver Pickford

Return to News & Insights