AI: The new battleground? – By Amy Smith

Introduction

We have all heard about the impacts of AI on the number of Tribunal claims and on certain applications. Documents which have normally been a few pages are now in the hundreds. There can be no doubt that AI can greatly assist litigants in person during proceedings to provide clear and detailed answers to questions, particularly in an unfamiliar legalistic environment. However, my experience of litigants using AI has, on the whole, only led to more confusion and increased costs. The big issue for the system is perhaps the scariest: can the overworked, and often argued underfunded, Tribunal cope with increasing demand? 

Witnesses are being taken to their witness statements in cross examination when their oral evidence deviates from it, to only reply: I used AI to draft the statement. It is clear that many litigants use AI to draft documents, but do not then critical analyse and engage with its content before sending it out. 

Litigants are making numerous complex and argumentative applications before the case has even been case managed. It used to be rare for a 30-page application for specific disclosure coming before standard disclosure has been ordered. It is now common. The application is highly likely going to be refused, the costs increased, and the litigant confused as to why they can’t have what AI told them they should ask for. 

After case management hearings, a litigant is asked to clarify their brief claims in the ET1. They then proceed to send a document hundreds of pages long. Not only will 90-99% of it potentially require amendment applications, the clarity sought will often be missing. The other side will have also had to spend hours carefully reading the document and cross referring it to the ET1 to even work that out. 

These are only a few of the many costly complications that AI can add to proceedings in the Employment Tribunal. 

So what can we do about it? Frustratingly at the moment it feels like very little. It seems clear that future changes to Tribunal Rules and Presidential Guidance will be needed. Below is a summary of a few tips, methods and approaches that could help to manage these types of cases. I’m hoping that this article can start as a jumping off point for practitioners to share their tried and tested approaches. 

Ideas

To flag or not to flag?

• We all know that feeling of dread when you see the Particulars of Claim stretches over a considerable number of pages with no clear indications of the actual claims. 
• In the response, it is of course open to a respondent to flag that the claimant is clearly using AI, and highlight that the claimant is encouraged to review and condense any AI produced document before sending it on. This will hopefully draw a marker, which can be returned to later in proceedings if the claimant continues with this approach. 
• The risk here is that it will either be ignored by the claimant, or it will encourage further anger from them. 
• However, including such a statement using wording that doesn’t rebuke, but rather warns of the risks and encourages a wise and critical approach to AI, may give the litigant pause for thought.
• Primarily, it could help later on down the line with any strike out or costs applications due to repeated and unreasonable use of AI.

To respond in full or not?

• When in receipt of the lengthy AI drafted ET1, the temptation is to respond in full. We practitioners never want to be accused of conceding something by our silence. 
• However, responding to every element and fact within an AI monologue is unlikely to assist the Tribunal. It is likely to keep the argument within the claimant’s arena, instead of one crafted to assist both sides and the Tribunal. 
• My recommendation would be to focus on your client’s story of the employment relationship, keeping aspects to key factual headings, whilst also attempting to focus on the potential legal questions.

To read or not to read?

• When asked to clarify their claims, the litigant in person sends you a lengthy document worthy of competition with The Odyssey. Our first instinct is to sit down with the highlighter and work through it as best we can. 
• There is nothing stopping you from writing to the litigant and informing them that: 1. The claimant was ordered to provide clarification of the existing claims, 2. The lengthy document on a cursory read does not do so, 3. You will not be reading the document in detail, 4. The claimant is to comply with the order to clarify and you expect they will do so by {date}, 5. You recommend that the litigant read their own document and condense it so that it provides the information requested. 
• This type of further and better particulars cannot be acted on until a Preliminary Hearing is listed. These hearings are often listed for 2 to 3 hours. At that point you can raise with the Judge that the document is too unwieldy to be gone through in detail in the hearing, and disclose the requests you have made for the litigant to comply.
• The risks with this approach are that the litigant will simply ignore you or send even more irrelevant material your way. Further, the Tribunal is unlikely to look favourably on a represented party refusing to read correspondence without good reason. Finally, it will mean a wasted Preliminary Hearing. 
• The benefits of this approach are costs and time saving. In addition, it may cause the litigant to reconsider their voluminous approach. 
• My view is that this is a perfectly admissible approach in these circumstances, as long as the further and better particulars are incredibly lengthy.

To strike out or not to strike out?

• If a party is repeatedly using AI to produce unreasonably lengthy documents such as applications and witness statements, there comes a point when a fair trial may no longer be possible. This approach has existed way before AI was even a twinkle in a programmer’s eye. 
• If the party is refusing to engage within the four walls of a claim structure and will not be tied to a list of issues, an application to strike out under grounds (a), (b), (c), and (e) can be made. 
• This is, of course, something of a last resort. I would only recommend this approach once multiple hearings and attempts have been made. It will also be incredibly fact specific. 
• Another benefit of this approach may be that either the Tribunal or the litigant is pushed into managing the case appropriately going forward. Whilst the Tribunal does not, yet, have any explicit powers to manage the usage of AI by a party, strongly and clearly worded case management orders (such as unless orders), can be the encouragement a litigant needs to engage properly with the information they are providing.
  

To challenge or not to challenge?

• A big early theme of the use of AI when drafting skeletons focused on represented parties including fake case names with made up principles. We have all read what happened to those practitioners. 
• There is no doubt that, should a regulated solicitor or barrister provide a skeleton or submission to the Tribunal based on false information, consequences can be severe. 
• However, with litigants in person it is a common occurrence. So should practitioners be highlighting these fundamental misleading statements to the Tribunal? 
• My experience is that it is not worth doing in any detail. When I have raised this in the past, Tribunal Judges have said that they know what the law is, and they will be careful when reviewing both sides’ skeletons. But they do not wish me to go case by case. 
• My recommendation would be to flag it to the Judge with a headline point, but not challenge each particular case with your findings. That way, the Judge does not fall into error and time is saved. 
• The other angle to this is, however, witness statements. Once a witness confirms they have read their statement and sworn/affirmed the truth of its contents, it is, in my view, entirely appropriate to openly challenge any inconsistencies in cross examination and submissions. 
• Whilst a skeleton contains argument, the witness statement should contain facts. And the swearing an oath/taking an affirmation on its contents should be taken extremely seriously. 
• When someone uses AI to create a witness statement, it is automatically being written in someone else’s voice. Further, AI is created to be helpful and may fill in some of the gaps. Some litigants read the statement with a cursory eye and, because it uses legalistic or complicated language, they feel it makes them look a certain way: cleverer or more educated. So they accept its contents without digging into the detail. 
• Whilst it is true that simply because the witness has got one thing wrong, it does not mean all their evidence is automatically tainted. However, it still presents difficulties. How can a Tribunal properly do its job assessing the evidence if a statement contains matters which are not true, and the witness hasn’t realised they are in there. How much more of the statement is affected by the litigant’s preferential bias when reading it. 
• From experience, Tribunals are open to hearing arguments on this point. And I have had success in persuading Tribunals that this has infected the evidence as a whole. 

Conclusion

The above are just a few of my thoughts as to how to try and manage the growing dilemma of AI. I’d be grateful and encourage any comments or stories, as it would be great to hear from other employment practitioners who have had similar experiences. Sharing the knowledge can only assist the overburdened Tribunal system by freeing up the time needed to manage and hear these important cases.