Risk Rewired Episode 9

Speaker

Transcript

Gregor Hayworth (Director)

[Music] Welcome to another episode of Risk Rewired: the Energy Disputes podcast. I'm your host Gregor Hayworth, a Director in Burges Salmon's Energy Disputes team and I'm very pleased to be joined today by Burges Salmon's own AI guru, Tom Whittaker. Today we'll be discussing the use of AI in energy disputes, rather than disputes arising from the use of AI more generally in the energy sector.

So welcome Tom, before we get started would you like to introduce yourself?

Tom Whittaker (Director)

Hi Gregor, hi everybody, it's great to be here, thanks for having me. So I'm a Director in Burges Salmon's Technology and Dispute Resolution teams, and I advise, train private and public sector organizations on AI regulation, legal risk, and governance and I write and speak on the topic as well. But importantly for the last 10 years I've been using AI as part of litigation practice to be able to run cases more efficiently and effectively.

Gregor

Thanks Tom. Well I thought it might be helpful maybe just to kick off with a fairly basic but important question, the term AI it's a real buzz word that's been used over the last few years more and more but what do we actually mean when we talk about AI?

Tom

Thank you. It's a great question. It's always important to go back to these foundational points in each case where we're looking to potentially use AI because everybody has a different understanding of AI and how it's deployed in practice in a particular case, or by a particular vendor or organization is likely to differ as well. So when it comes to AI, plenty of different definitions but you can see certain definitions coming through within legislation, regulation, and guidance as well. So if we take for example the EU AI Act, which was enacted in mid 2024 and which is has a number of transition periods and is start- the provisions are becoming enacted throughout 2025 onwards. That defines AI as a system which is machine-based designed to operate with varying levels of autonomy and that may exhibit adaptiveness after deployment, and that for explicit or implicit objectives infers from the input receives how to generate outputs such as predictions, content recommendations, or decisions that can influence physical or virtual environments.

So a fairly lengthy definition. But what you can see from that definition when you compare it and contrast it to other definitions of AI, the real concern with AI and why it's special regulatory or policy treatment is that there's the potential for autonomy; it may act separately and differently to how humans want it to. It may adapt in that it may be learning over time and changing what it's doing and how it's doing it. That it's able to generate various forms of content whether that's changing text to video or audio, or whether that's changing video or audio to text and that ultimately it can influence the environments in which it's in, and that's either because it directly results in a decision or outcome, or it's an outcome which a human then needs to make use of and properly understand.

So those are some of the risks and issues around it, but what I would say to everybody listening is that whether you're a lawyer or whether you're a client you need to ask your vendor, your organization or those who know, what is the AI system that you're intending to use and how is it been implemented in your specific circumstances?

Gregor

Excellent, thank you very much. And I think, am I right in saying you have prepared a helpful glossary or document to help possibly our listeners understand some of these AI terms?

Tom

Thank you for the plug, I appreciate it. So Burges Salmon has an AI law regulation and governance blog and one of the many things that we have on there is a glossary which then goes through a series of key AI terms and then it shows you exactly how they've been used in law, regulation, or policy, and then it has the hyperlinks to those as well so you can go and see the underlying source document too. But hopefully that's a useful tool that people can use if they want to see how these terms are actually used in practice.

Gregor

Perfect. So Tom, given that there clearly is an increased use of AI generally speaking, but specifically in relation to energy disputes and litigation. Do you think that that will mean all lawyers moving forward are going to have to be AI experts?

Tom

It's a good question and one that I know plenty of lawyers are thinking about and will continue to think about and it's not necessarily a case that they have to be AI experts but it is that they have sufficient, what's called, AI literacy. And so there's two reasons for thinking that. One is that lawyers are under regulatory obligations to keep up to date with legal developments and legal practice and then to act in the best interest of their clients, there's a minimum level of understanding that they need to have. But it's also because you can see within the wider regulatory and policy world when it comes to AI, discussions about obligations for AI literacy. Now for take for example the EU AI act which has specific obligations for AI literacy. You've got to be quite careful and thoughtful with exactly how those apply and they don't necessarily apply to all lawyers in all circumstances, so you've got to think about who needs to know what, when, and to what level, and the level of understanding and expertise they already have. And so for a lawyer who is thinking about making daily use of AI, they will have to have a different level of AI literacy in those circumstances tailored to those circumstances which are different to someone who is designing the overall system, testing it, verifying it, and then explaining it to the clients or courts.

So I would say rather than think about AI expertise, think about AI literacy and thinking about what would apply in your circumstances for what it is that you want to do.

So Gregor, I'd like to think about how we can make use of AI within litigation and so I'll need to think about how we can tailor it to the specific circumstances of your cases. So what I'd love to hear from you is the nature and dynamics of energy disputes and maybe some examples of the types of disputes that you've been involved with.

Gregor

Certainly. So I guess generally speaking energy related disputes love an energy producing asset or assets at the root of the dispute, so that could be an offshore/onshore wind farm, a solar farm, power plant battery storage system. So essentially the dispute is going to revolve around the technology which could be technically complex, in terms of how that technology is designed and constructed, or how it's operated for example, and that is obviously key in allowing kind of optimal energy production to achieve what most energy assets are designed to do which is generate revenue. And so we also frequently see energy disputes which have- where the developer has essentially implemented a new and potentially untested technology and that can obviously add to the technical risk associated with those energy producing assets. So it's technically complex a lot of the time and quite an innovative field.

The energy assets themselves are generally expensive to build, or purchase if you are purchasing from an already existing asset, and often to realize a satisfactory return on investment those assets need to be generating energy for many years. So that means that energy disputes can often involve long-term contracts, could have a joint venture agreement for example, or party a club together to buy the asset or operate the asset, an operation and maintenance agreement for example, or a per purchase agreement. So long-term relationships often are a key theme in energy disputes, which adds to the kind of intricate tight rope that needs to be walked when preparing disputes and thinking about strategy.

Energy projects, they also involve multiple stakeholders at various different phases of the project life cycle and that could be funders, investors, the supply chain, or off-takers of the energy being generated.

The projects also tend to have some form of international element to them. That could be specialist contractor manufacturing items which are based abroad, or materials that need to be sourced from other jurisdictions, and that cross-border nature can often lead to difficulties in the delivery of the project. So we saw relatively recently the incident involving the cargo vessel the Ever Given, which caused a massive blockage in the Suez Canal, and that resulted in a massive- or this resulted in a major backlog of vessels being unable to deliver goods and materials, and many of those goods and materials were to be used in the construction or operation of energy projects.

So with all that generally means energy disputes will be document heavy, they will usually involve a significant number of witnesses, and whether that be factual witnesses or expert witnesses, and generally speaking they will also involve multiple data sources that will need to be considered.

Tom

Brilliant, thank you Gregor. What I'd really appreciate is if you could maybe give an example of a dispute that you've been involved with which can bring some of those dynamics from energy disputes to life.

Gregor

Of course, so a recent example of an energy dispute I was instructed on related to the construction of an offshore wind farm which had run into significant difficulties during that build. The crux of the dispute was that the main contractor who was tasked with designing and building the wind farm was claiming it was entitled to a significant additional sum of money and a lengthy extension of time which my client, who was the developer of the wind farm, didn't agree that that contractor was entitled to under the contract. So that dispute touched on almost every phase of the project, from manufacturing of the foundations, through to transportation and installation of the wind turbines themselves.

And it also covered a period of a number of years, so it was very complex technically. It involved multiple jurisdictions and required the collection and assessment of a huge amount of technical project documentation, and communications between the various stakeholders and parties involved. There was a significant number of factual witnesses that we had to take evidence from, and we also required to instruct and work with various expert witnesses, there's a number of technical expert witnesses, and there's also a delay expert and a quantum expert as well.

Tom

So that was a really interesting example of an energy dispute and so what I'm hearing from that when I'm thinking about how I would run a case and the different stages to it, is I'm hearing that you've got those multiple different stages; disclosure, factual witness evidence, multiple stages of expert evidence. Consequently, you've got different data points from different data sources, potentially managed by different parties or different parts of the organization. So there's questions about how you get access to it, how you forensically collect it, and then ultimately how you're able to review it in an efficient way so that you really get stuck into the detail but also so that you're able to analyse it at a higher level so that you can see the bigger picture. And you'll be thinking about that from the outset of a case because you may want to start doing some early case assessment to get a direction on where the case could be going, to inform strategy, and to inform your thinking about the project management. But it's also then thinking about how you can make best use of that data at every stage along the way, so that you can run each stage efficiently and effectively. Am I hearing the sorts of things that you were thinking about? Does that sound like I'm putting myself in your shoes?

Gregor

Absolutely, yes. There was a lot of planning and project management of the various different work streams to ensure that we were progressing the matter efficiently and covering all the bases in terms of recovery of documents, obviously to ensure that we were getting the full picture and preparing the cases as thoroughly as possible.

We've discussed obviously the kind of greater use of AI more generally but also in litigation, if we had a similar dispute, this year for example, there's obviously a greater access to the use of AI Technologies, what differences or what opportunities might there be and what should lawyers and energy companies be considering when thinking about using AI?

Tom

Yeah, I think let's start by just putting this in a little bit of context. So we've had more than 10 years of using forms of AI within litigation and then some of the earlier forms I think go back into the late 20th century, but in particular the use of things like Technology Assisted Review through the mid-2000s and then the likes of Continuous Active Learning and predictive coding through the late 2000s and then 2010s.

And so, there's always been access to some form of AI systems but they've been traditional AI systems where you have to get a third-party vendor and they're usually for the bigger set pieces, such as disclosure. What we're really seeing now over the last few years of the early 2020s is the rise of generative AI, because of the Large Language Models, and in particular because you have the competition in the different Large Language Model providers and that they are able to give access to other parties who then integrate it within their litigation focused AI systems, or other systems.

So you start to see GenAI enabled disclosure providers and platforms, or you start to see start-ups who are making use of GenAI for specific litigation focus tasks. And people may, who are listening, may have seen that we've announced that we've been partnering with a company called Wexler who have been using GenAI to help build out chronologies, but who over the last few years, whilst we've been working with them and we've been giving feedback and working closely with their development team, we've seen them develop their litigation focus product, using GenAI, to be able to extract data, summarize data. All of this with the ability to then look exactly at the citation and the source of the documents, for the humans to then verify it, and then to be able to use that data in efficient ways however they need to within the case.

So where I would say the AI opportunities are in the future, well there's multiple but broadly go into a few different categories. One is obviously disclosure, as much as technology is coming along to try and help deal with ever greater amounts of data that we need to get through, the world's data landscapes are increasing so energy companies and providers who are listening they should be thinking about their data and information governance, what data they hold, where it is, if it's in a structured high quality format. Because ultimately they may need to get access to that at short notice to be able to respond to Regulators or to litigation as well.

The next bit would be, well, when you get through to disclosure then you're making use of AI in different ways. Now some of that may be behind the scenes, you can use AI to help tidy up OCR recognition of handwriting or other annotations in hard copies which you're scanning in or have already been scanned in. But you also may be using AI in other ways to be able to categorize or classify documents. So for example automatically identifying this is a contract, or this is a building contract, or subcontractor agreement, or something like that. Or to automatically extract clauses from those if you want to quickly hone in on the Force Majeure Clauses, for example, if that's what your dispute is about.

And then ultimately you've got potential use of GenAI for saying whether a document is relevant or not and then it's up to the lawyers to think about how they put that within their workflow. Does that help with their first tier review, to be able to identify potentially relevant or not? Does that instead act as a quality check, so you have your human review but then you use GenAI to act as a sense check to help identify any outliers, or points that need to be checked? So that's very much disclosure. The great thing with GenAI is that you can use it in other ways as well.

The great thing with GenAI is that you can use it in other ways as well. So we see it from some providers who are saying that you can use AI to help marshal the documents and get earlier insights into what they say. For example by building out Dramatis Personae by building out chronologies helping you interrogate the facts and the chronologies in greater detail, always going back to the source so that you can then understand properly the context, potential reasoning behind the extraction.

And then you also start to see a number of vendors producing AI tools which help with developing case strategy. So you put your documents in there and you say can you show the person X, for example, knew issue Y. Or can you see anything which contradicts my argument here or my assertion here. And that helps you start to test things out based on the underlying data. But to go back to my early point, this is why data is so important, this is why the listeners data, or the way, if they're lawyers, the way that they collect data is so important because it's that that the AI system then has to interrogate. And as much as gen AI is great with large amounts of unstructured data, with natural language, you as a lawyer, if you're the lawyer listening you still then have to make use of it within the AI system to then be able to put it into an output that is useful as part of your analysis. So you need to make sure that the metadata is there, that the labelling is there, so that you know what documents are in there and importantly what documents are not in there, so that you're not slightly skewing the results. Data, data management, data governance are going to be critical to the use of AI within all this.

But ultimately, to go back to your question Gregor, the opportunities for AI are at multiple stages within disputes, not just disclosure and they're also for those sorts of frontend big ticket bits, like reviewing for relevance but also for key stages which enable those big bits as well. And so really there's plenty of opportunities, the difficulty will actually be working out which opportunities to focus on and then how you exploit them.

So Gregor, after everything we've been discussing today I'd love to hear from your perspective what are the key takeaways about potential use of AI in litigation?

Gregor

So I think for me the specific or key takeaways are, why are you using AI in the first instance? I think that needs to be at the forefront of the lawyer's thoughts and the client's thoughts. And I think an assessment of the risks of using AI is going to be key as well and particularly in respect of the specific case that it's being used, you'll have to make a decision on a case by case basis, not all disputes are going to lend well to the use of AI. And I guess you also need to consider how AI fits within the wider work streams of the case, preparation, and what outputs the AI is going to generate and how that's going to benefit the case as a whole.

Tom

Thanks, I'd agree with those points and from my perspective the key point is to understand your and your clients objectives, the 'why' as you mentioned. And that's because AI tools have great promise in the right circumstances to really help on efficiency and effectiveness and to deliver against a series of different factors at an earlier or different stages of a case that otherwise previously we've just not had as much of an opportunity to do.

However AI tools are not a Panacea, they come with risks and they do need to be carefully considered in the circumstances as well. And also fundamentally they aren't necessarily the right tool for the issue or problem that you have in mind, and if I can leave with one final thought on that I often hear about people saying that they use AI to help summarize emails from lawyers. If anybody ever has to summarize an email from me using an AI tool I would much prefer that they just asked me to write a shorter email or to explain why that level of detail was necessary. The AI tool is not necessarily the answer, there may be a better way of doing it but it should still be considered.

Gregor

Absolutely, well thank you thank you very much Thomas. It has been extremely interesting and I've enjoyed our discussions so thank you for joining me on this Energy Disputes podcast.

Tom

Thanks Gregor, thanks everybody for listening, it's been a pleasure to be here. [Music]

Gregor

Thank you for listening to this episode of Risk Rewired: the Energy Disputes podcast. If you'd like to know more about any of the points discussed on this episode or our Energy Disputes team more generally and how we might be able to assist you can contact us via our website. Thank you very much for listening. [Music]