AI Reckonings Risk, Rights & Resolution

[music] Well, good morning everyone. Um, that's going to be a tough act to follow, but we will try our very best. Um, uh, I just wanted to introduce this session by saying that we're looking at, um, AI, um, in the disputes world. Uh, and by that I mean we're try we're distinguishing it from how AI can help disputes lawyers in their work. If we have time and I think we probably won't, we might come on to that subject at the end.

Um, but what we're really focusing on is um is how AI impacts um uh disputes uh and how uh the law can adapt and change uh to meet the challenges that AI brings. So just to frame the discussion, I thought it'd be quite interesting uh initially because we hear so much about AI. It's always a topic of discussion amongst friends around the dinner table in restaurants. Um but what is it that we mean uh about AI in a legal context? Um and in order to sort of frame the challenges I think we're we're facing um I think an initial question I'd like to ask the panel um is what do they understand?

what do what do they see as uh uh uh as AI? What do they understand by AI? Because they all work, they're all very experienced, very distinguished. They work in the AI field. Um but I think we can all benefit from their um their knowledge.

Um and um and their experience of of working with AI. Um so let's start with you DA if we can. Um so you're deputy general counsel at G42 which is uh an exciting Abu Dhabi based company um that has a mission uh to build what it calls the intelligence grid um already it has 10 companies in its portfolio um and whilst been waiting I've heard how exhausting that's been as deputy GC acquiring those companies um but what what are they doing what what's G42 doing uh and and how is it um using AI in its in its portfolio companies.

Thanks Graeme. Uh yes, so we are doing a lot indeed. Um at times very hard to keep up, but um for us maybe I'll just explain actually what the intelligence grid is initially for for those that that may not know, but it's essentially a a concept that embodies the vision our group CEO at G42 has, which is the vision is that AI will become almost like electricity in the future. That's the way that he sees it. It will become readily available and it will infiltrate our daily lives and everybody should have the right to access it.

And so with that vision in mind, G42 has sought to create or establish a number of portfolio companies. I think we are at 10 now if I'm not mistaken. And all of them um cross multiple sectors. So for example um we have CPX which one is one of the more recent subsidiaries and that focuses on um cyber security. We have um Kasnner which is one of our joint ventures between G42 MGX and Silver Lake and that's really our data center business and then we put core 42 on the top which is the infrastructure layer.

you know, they're the ones that are putting in the servers, the GPUs in the data centers. We also have inception which focuses on advanced AI models, your large language models. Um some of you would have seen you know the announcement that we're looking to do collaborations with open AI and there are several several other portfolio companies across different industries including M42 which of course is the joint venture between G42 and MADLAR and that focuses on combining G42's digital healthcare and and MAB's hospitals. Um, the list kind of goes on, but all of these portfolios are kind of established with a view to ensuring that AI spans multiple industries and sectors, but connect in a way that, you know, we can benefit from AI in our daily lives.

Magda, that's interesting. And Magda, you're um the general council of AIQ, which is more focused on the energy sector. Um, what is it that AIQ is doing um or looking at in the AI space? what aren't we doing? So, actually, I love what DA was saying about uh, you know, AI being electricity because what AIQ is doing is actually we already look at the symbiosis between energy and AI.

So, we're a joint venture between Adno and Preight, another one of the G42 companies. So essentially when I when we think about AI we're taking data we're taking insights of years and years of activity in the oil and gas and energy sector from from adno NOC's IOC's and then we're combining them with supercomputing capacity um from uh the G42 group to create AI solutions for the energy sector. So in reality what does that mean? It means that we're utilizing those data insights to automate to bring efficiency to bring productivity to bring sustainability to the energy sector and and just by way of example one of our kind of key products that we have energy to to the power of AI which we've developed in alignment with inception Microsoft and ADNO um if you look at seismic interpretation um an activity that did take for example months and months we now utilize our seismic agent to do in the matter of minutes and hours.

And and Danny from a I mean it it's interesting I I call them startups and that would be probably slightly offensive. They're not startup companies. They're now well established companies. But um AI is also used um isn't it in in established uh companies globally. It's not so that uh there are companies obviously that are focusing on AI hate the word but I use it solutions um to sort of uh foster and generate sort of efficiencies in their own spheres.

What what are you seeing in that uh in that space with your sort of client base?

Sure. Uh happy to address that and first let me just thank our hosts. You you cannot do AI and not visit this region. Um, I've been here six times in the last few months uh to the Gulf and the G42 and AIQ are two examples of the excitement of what's going on here. So, what are companies doing?

Companies are dramatically rushing to implement generative AI across their business structures. Um, my my practice really divides into two. I've I've got the innovators, so the um LLM large language model makers of the world, but it's also the

what? Sorry to interrupt. What's large language?

We'll get to that. We we'll discuss that. That that that's a great question and it gets to the legal liability issues. We're going to talk about LLMs. Just a quick is those are the engine behind the chat bots that you're all familiar with.

So if you've spoken to chat GPT or Claude or Gemini or some of the other foundation models, those are powered by large language models which is the new version of artificial intelligence that's based on human language and can talk to you and you can talk to it just like you would a friend or a teacher or a student. So that that's really the engine driving. And the key is language because that's what you used to have to be a data scientist to implement these tools throughout companies. Now I have Fortune 500 companies where every employee is their own data scientist because they can speak to these agents to these large language models and create their own internal solutions to business problems. That brings great opportunity but it also brings incredible risk to the integrity of information throughout the organization.

So I've worked with six of the fortune 10 on multinational governance programs to make sure that AI is deployed in a way that is additive to the corporation and doesn't create embedded issues. And it is a big project because you've got global regulation, you've got increasing global litigation, but most importantly you've got a technology that's new and people are still figuring out the best ways to implement it effectively. I I always say focus on ROI, but responsible AI drives ROI. So, it's part of innovation.

And Manesh, you're um uh you and I discussing um earlier some of the sort of legal challenges. So, before we move on to the sort of disputes world, um why is it um do you think and I I' I've split this into three separate questions, but why is it you think that AI is likely to create disputes? I mean on the one hand it sounds quite obvious that it will um and on the other um throughout time um and I'm older than a lot of people in this room but we've had new technology um throughout time and you don't hear as much about the possibility of disputes with developing technologies as you do with AI. Why is it specifically with AI do you think that we're we're staring at dispute scenarios?

I think there are probably three reasons. I I think the first is scale. We've already heard from the panelists how much AI is being used. It is going to be everything everywhere as the term goes. I think the CEO of IBM a few years ago said that every company is going to be an AI company.

So it is going to be a ubiquitous technology and and the way to think of it is as the way that other technologies are being powered. So it's going to be everywhere. So the scale of it means it's inevitably going to create legal issues. The second I think is a unique feature about AI compared to other technology which is how it operates. If you think about conventional software, it's worth thinking about that in the sense that it's automated.

It follows human-based rules. It doesn't operate autonomously. Conventional software can't operate outside of the bounds of the rules that humans have created. The essence of AI is the opposite of that. It's to give computers software the autonomy to make decisions for themselves to produce independent output.

That means that AI is unpredictable, unforeseeable and it means inevitably that there are going to be unintended outcomes and of course that generates disputes. And I think the third reason is that as we'll come on to discuss our liability frameworks that we have currently in the world in all jurisdictions have been developed with humans in mind. When we think about concepts such as foreseeability, causation, which drive the disputes world, drive where liability sits, they're all about humans doing things or not doing things. And when we're dealing with now machines that can operate autonomously, we don't yet have mature enough legal frameworks to be able to deal with how we as humans think liability should sit in that environment.

And that kind of brings us on, doesn't it, to regulation and the need for regulation. Um I mean h how are and and there's been some recent developments particularly in the EU and we can talk about the US as well. I think that will be interesting to see what the the differences and the challenges are. But um how are regulations dealing with um the sort of AI liability uh space and and do we need specific regulation or we can can we make do with a sort of an exanti bolt-on to existing regulations? So I think and obviously welcome views from the other panelists um and I work a lot in the EU and the EU as we all know is an environment which is heavily regulated and certainly tech regulation is coming thick and fast.

You probably all have heard of the EU AI act which is effectively the EU's answer to AI similar to the GDPR in terms of its impact and scale. The way the EU is dealing with this is to implement what I would call X anti-regulation. It's about trying to prevent the harm that AI could cause earlier on. It's not about responding to the harm through a liability framework. And I think a lot of AI regulation that's coming through, including in the US, including in Asia, is all about targeting the development andor the use of AI.

So trying to tackle the problem at its source rather than trying to provide mechanisms for redress afterwards. And I think that's the holy grail really actually is how you deal with the expost aspect of of AI. When things go wrong, how do you do that? That's challenging. And I think that's part of the reason we're seeing this X anti-regulation, implementing requirements for governance, making sure it's developed and used in the right way.

And how uh and I'll come on Danny to the US uh before we talk about uh what's happening locally, but um h how how does that work in practice? Can you give us an example of how the EU regulations are trying to grapple with that?

So, one challenge that the EU's had to face and all regulators have to face this is do you regulate the technology itself? Do you regulate the way it's used or do you regulate outcomes? Now, a lot of regulation we have currently is aimed at humans and it's aimed at outcomes. We say don't cause discrimination as humans. But I always say you can't rewire human brains very easily, but you can rewire machines.

So AI gives us the opportunity to regulate the actual technology itself. It can say make sure you develop that in a way which doesn't cause discrimination in the first place. But the problem is if you regulate too heavily on the technology itself, you stifle innovation. And actually some of the harms caused by AI are caused by the users. So you've got to find this tricky balance between regulating the technology and the use and that's something that regulators are struggling with.

So the EUA act actually does both. It principally targets use of AI and it has a high-risk use regime but it also targets the development of the most powerful models which are called general purpose AI models under the EUAI act. So foundation models, frontier models, LLMs that Danny was talking about. And Danny, is that is that a similar path? Is that is that a reflected in the US regulatory regime as well or are they approaching it a different way?

No, no, it's quite the opposite. And and I would take a step back and say I I take a fairly global view to AI regulation. One of my mandates is as senior consultant to the UN on parliamentary engagement for artificial intelligence. And and the goal there is can we come up working with parliamentarians from several dozen countries on harmonized standards for AI. And the reason is AI is not a jurisdictional issue.

Um that the same reason that David Saxs in the US last night tweeted about jurisdictional patchworks in the US. We don't want 50 state laws of AI. We also don't want so many nation national national laws on AI that we can't have the type of cooperation we need um when the US and the Emirates work together on very very important AI projects. So I I think less is more on AI regulation and we really need to focus on what the grave arms are. Um let me say this, the EU AI act I I have a lot of respect for the people who worked on it.

Um I I on behalf of clients was engaged in working on that uh law but it's been very controversial and I can tell you the private sector has made decisions on where to innovate based on that law and the fear of some of the administrative burdens that it creates a priority the the x x anti x x anti rather than expost. So regulation does have consequences. Um e even to the point where now I think the EU is is rethinking some of the deadlines and some of the uh impacts of that law and um there's some proposals along those lines. Um I would say we we know what the most serious harms are that AI can cause and I think countries need to work together on those. Um the rest I think a lot can be handled through laws of general application and the tort system.

Um, I agree with Nesh. I think it's a good point that there's a lot of new challenges to applying old laws to AI, but at the same time, I think it's going to be a fool's errand to come up with a law of AI in advance. I'll give you one example and then I'll I'll pass, but um I was on a panel uh in Geneva and and there was uh one of the authors of the EU AI act and they were asked to give an example of a lower risk uh use of AI and they said chat bots. Well, if if you look in the United States today, all the federal regulatory, all the state regulatory, all the litigation focused on chat bots. Chat bots, it turns out, can do just about anything.

So, we're not very good at predicting where the technology is going in advance. So, I think we need something like a common law that's much more adaptable.

And we'll come on to talk about that because that that's obviously one of the features of common law. But just sticking with regulations for the time being. Um uh Magda and DA, what what uh what are you seeing in the Emirates in terms of the regulatory landscape and um do you think that um uh that we should sort of hold back slightly and see how things work out in in Europe and the US or is there a is there a drive to our own uh domestic regulatory regime?

Yeah. So what I think is has been fantastic about this region generally is that actually since 2007 uh uh sorry 2017 when they appointed the AI minister here this has been a conversation that's been ongoing. So before open AI, before Sam Alman, before this was a conversation in the EU, this was already being, you know, a conversation, the opportunities, the challenges were already being looked at here. And what that's produced is an environment which is really really pro- innovation, but also pro kind of testing. So does the uh approach that the EU are aligned with, does that work?

Does the approach that the US is talking about, does that work? Um so we're taking a kind of a speculative view and and kind of testing and actually there's a lot of regulatory sandbox in in operation and in discussion in this region which I think are really really helpful and will align with the the point that was just made in terms of you know we don't know exactly the the harms that can be caused and the speed of the development of this uh this technology is so fast we need to move at the same pace. I think what's also fantastic about this region kind of just just generally and even just looking around the room, it's a it's a testament to the collaborative approach and the um the the cultural the language diversity that we have here. Some of the things that we're talking about when we look at um you know global approaches to regulation and specifically in relation to AI is is how can you do that? How can you apply that to such a a broad range of interests to to cultural and and language diversity which sits in such a a breadth and and a um an interesting the interesting world that we live in.

So so that's a challenge that the E sorry the UAE is continually looking at as as well.

And I think um and do you subscribe to the view that an overregulated environment stifles innovation? I

I do actually. Yes. Um we are very much pro- innovation and often you know at G42 we need to move very very quickly. So as as Magda alluded to and the other panelists you know the technological advancements are are kind of moving at such a fast pace the regulation can't keep up. So as lawyers in house it can be challenging as well as to how you can advise the business.

That said, I do think there's a benefit to looking at the, you know, regulations that are in existence now. And I do wonder if the EU AI act, whilst I think, you know, I have some issues with with with the act and how it approaches a AI, will it become a little bit like the GDPR in the sense that will we start to look at it as an international guide? Um, and I can already sense internally at G42 that we are, you know, very much focused on the EUAI act and we're actually in the process of doing a regulatory analysis to understand how those regulations could could touch upon our business, particularly as we're looking at EU expansion. Um, but we've also um set up an RAI function that's dedicated to this topic. And of course very important for us that when we're deploying or developing products that we're looking at, you know, the safety of these products and and thinking about ethics.

And so as part of that, we're not just looking at regulations, we're also looking at international standards and frameworks. Um for example, the ones that we are very much focused on are the the NIST AI framework. We're looking at the OECD AI principles, the G7 Hiroshima International principles. So I think there's enough out there that can help guide us in the interim.

But fundamentally I agree. I mean um I remember when I first came out we looking at financial regulation uh and everyone looks and tries to get a best-in-class regulatory regime based on global standards but they're not always fit for purpose in the region because you know Abu Dhabi is a leader in AI innovation. Um and again you know don't want to stifle innovation uh and some of these regimes are just not suitable. I think that's that's probably where we land. Um, and going back to the common law, so what else is it that just leaving aside the sort of statutory and the regulatory regime?

Uh, and we're touching on Danny, I think we're coming on to the subject of of of what what you were speaking about earlier. Um, what else are lawyers doing to sort of foster this sort of, you know, uh, new AI environment? Uh you know we talked about the laws of negligence and the laws of contract and in the common law for example um you know for those of you at civil law train we call it judge made law um and for many many um well couple of centuries at least in the UK um it's adapted to new technologies and you know the common law is very adaptive u why why should we not just leave this um to judges?

Yeah,

that wasn't a rhetorical question by the way. [laughter] Was that

Danny? You Yeah. Sorry, that one's for you because I know you wanted to talk about this.

Well, I I love I mean, this is my favorite topic, you know, 10 years ago. Um I I was a much younger lawyer and um working on technology litigation and my my concern was this was back when our technology was dumb and it was still too hard and too expensive to litigate and all the seeds of what we're seeing in the disruption of the law today were apparent back then. There w there was a case on unintended acceleration in uh motor vehicles and one of the opinions the judge said you know what I've looked at all these expert reports and it's just too complicated a pox on all your houses and I I started thinking about that and I wrote a paper that was controversial at the time but I said this is going to flip the whole system of liability because it's going to be too expensive for any plaintiff to prove who caused What when you have large tech stacks, trillions of parameters, lots of different people fine-tuning, changing, implementing, agentizing, you know, how could any court system unscramble that egg? So, you know, those things have have in some ways come true. Europe has come out with a revised product liability directive saying we're going to flip the burden of proof, sort of guilty until proven innocent.

Um, I don't think that's a a great solution, but it's a very hard problem. Um, I will say I think and and and I do truly believe this, so many of the harms of AI can be prevented with easy things. And the problem is often that we get so focused on the fancy problems. You know, what if the next generation of of chat bots becomes general intelligence and does, you know, really bad things? That's a government level problem.

But people throwing out underested, underresourced, poorly designed products that are capable of making medical decisions, legal decisions, financial decisions. That's a problem we know how to fix. Uh we can fix it preemptively. Working with companies on risk mitigation, which is 60% of what I do right now, red teaming, uh bias testing, th those those are not hard to do. we can do them with legal privilege and get things fixed before they go to market.

Then if companies behave well and do those good things, which are not very hard, the rest we can sort out with litigation. So I think if we kind of don't get too distracted by the science fiction stuff and really focus on what could harm people and how do we stop that, I think we'll have some practical solutions. and Mesh, what are you seeing um in the market just in terms of disputes um that are are starting to come out of the the industry and the use of AI? Um what sort of disputes are you seeing in your practice?

I think it's interesting because I think we had a sense that with AI we'd get the sort of mammoth disputes which would go to the heart of who is liable when AI causes harm, but actually what we're seeing is sort of AI disputes spring up in various different legal areas. So I think the big area that probably most of you will be familiar with is IP and copyright. So there's a huge issue, a dichotomy between content creators and and their rights and developers and and their use of material to train models. And so that's played out in a few jurisdictions. In the US, in the UK, we've got a big case Getty and Stability, which you may know about.

In Germany, there was a very interesting decision just a few weeks ago, and there's no consensus on that. it does go to a sort of moral and anthropological idea about where we think um the right should lie when it comes to content. So IP is an area where this is playing out quite interestingly and then there are a few other areas where it's playing out as well. HR and employment I think is going to be a huge space where AI is going to cause some issues. There's a claim at the moment in the US, a class action against workday.

And the claim is essentially that AI tools used in the recruitment process have caused discrimination. And it's a fairly simple theory and you can understand why claims like that have arisen. So that's another area where I think we're going to see that. And we've already seen it in the US, we're going to see it in the EU, we've got it in the UK now. Class actions, they're becoming more attractive for claimants.

the damages figures that we're seeing are becoming mind-boggling and I think quite a lot of big classes are going to go after large developers for for AI disputes and then we've got some other interesting areas playing out one of them is this concept that I used to talk about it's coming back to the four now AI washing where organizations make false claims about using AI or the extent to which they're using AI and actually the SEC has fined organizations in the US for for AI washing And I think in the most egregious of cases, a company was saying they were doing all audits and tax returns via AI, but actually it was just a room full of people in India, I think, doing it all manually. So there are some really egregious cases like that. So we're starting to see different areas where this will pop up and I think we'll start to see more of those. And Danny, before I I'll uh and I'll come on to uh Magda and DA shortly, but Danny, I can't leave the subject of uh lawsuits and litigation without talking about what's happening in the States. Uh because that's always a bit of a crazy place for litigation.

Um I read something about um generative AI hallucination lawsuits. Um what are they? Well, so I I uh got to work on the first two of those and and and so I'll I'll speak not about any particular case, but at a high level, but I actually think these are the most important currently AI litigations to be aware of because hallucination is such a silly word. Uh we like to talk about how generative or large language models hallucinate. I would love to erase that word from from from the uh computer science language.

They don't hallucinate. Sometimes they're right and sometimes they're wrong. The way these models work, especially I'm talking about large language models. They don't have an inherent sense of facts, of truth, of right or wrong. They are large language models.

I said that earlier. The key word is language. They're trained on language and all they're doing is stringing words together based on the most common patterns in the training data. So if they tell you, you know, if you show it a picture and say, "Is this a cat or a dog?" It it doesn't know if it's a cat or a dog.

It's just looking at the images and saying, "Okay, that matches this pattern, so we'll string some words together that give you an answer." Most of the time it gets it right, but sometimes it gets it wrong. And what people have to realize is none of the companies are saying these are large truth models or large fact models. They're large language models. So if we use them appropriately and we ground them in facts through technologies that are available to make them behave responsibly then you can get to a product that's very effective.

So that comes back to litigation. Um there was a case it was the Air Canada case and uh this was in in Canada. It was sort of a small claims court but the outcome got global attention. um even though it was a really minor silly kind of case because a guy asked a chatbot uh you know can I get a discount on a particular type of fair and the chatbot said yes you can. Turns out that was false.

Um and the court said no. Air Canada, this Canadian airline is responsible and bound by what its chatbot says. Even though there were disclaimers all over the website that said it's a chatbot. Check check the facts. It even gave that person a link to go click on it and read the actual policy.

Had the person done it, the policy would have said no. So taking your point that sometimes US gets the uh image of crazy litigation. Actually, in the cases I worked on, they came out the opposite way uh in favor of the chat bots because um in in the one published opinion, the court said, "No, that the chatbot told you. It said this is an experimental tool. This is how it works.

It's not the final source of truth. Check your facts." So, it's, you know, we have laws to deal with that reasonable reliance. When something says don't rely on this, don't rely on it. Use it.

it's very useful but then go fact check that's the responsibility of keeping humans in the loop I actually think that should be a very optimistic point for everyone in the audience because you know Mark made a great point on the last panel we all fears feel some fear of replacement but at least in the current state of affairs humans have the most important role of all which is making sure that what the AI tool encourage uses us encourages us to do is right we're still the gatekeepers we still have that ultimate need an authority to check the facts. Um, that's going to inform the litigations on these uh hallucination lawsuits, but it's also going to inform safe and good uh humanentric AI.

And I should add, Danny, that view of crazy US litigation is my personal view, not Akim's view. [laughter]

And that goes for I'm sure everybody on the panel speaking for ourselves. So um so now moving on to uh Magda again and Donette um obviously in your institution your companies you're developing you're at the forefront of AI development um do lawsuits or the potential for lawsuits sort of keep you awake at night? What what are you doing to to mitigate the the potential for lawsuits in in what it is you're developing? Magdala, let's start with you.

Yeah. So if I may actually I'd love to draw on a couple of points that the the the panel's already mentioned. So I think responsible humans create responsible AI. The technology in itself isn't uh evil, isn't bad. So if we kind of work with the basis of um AI should be used as a an addition as an advisor to enhance the human experience, you know, uh we talk about AI for good and responsible AI.

So I think responsible people can achieve that. And to to add to a couple of the points that Danny was making, it's actually quite easy to do that if you take the time to kind of work it back and think about it. And actually, you can use a lot of your existing frameworks, existing policies, processes. I mean, how many of the businesses you work in already have quality management processes in place. If you look at some of the requirements of being a responsible AI provider, developer, and then some of the requirements of the EU AI act, they really build upon principles that businesses are already undertaking, already achieving.

And to add an element of um AI responsibility into that wouldn't probably take a lot of resource or time and effort once you've kind of figured it out, of course. Um so so doing things like uh keeping logs of your behavior, using quality data, the right amount of data um to train your models, um things like that are really really easy wins. You know, monitoring the the false positives, true positives from your models, um ensuring that um you're taking customer feedback. I mean one of the key things that that we've looked at is actually um creating an AI kind of native workforce internally at AIQ and then within our customer so that they have the appropriate skills to know how to use the tools responsibly and then also to understand how that they won't feel that kind of threat or to to use the tool as an enabler as I say for for AI for good for the for the betterment of their kind of work and and general life balance

and DA as your your company It has probably more of a global reach, doesn't it? With the number of companies it has now, um 10 at the last count, but dotted around the globe. Um is a sort of global environment something that's of a concern to you because don't you have to be on top of a global regulatory environment? And uh that adds another layer of complexity, doesn't it, to what you're you're pushing out?

Yeah, definitely. and and that's exactly why we've established this this responsible AI function that's dedicated to this and and that is going to be growing over time but essentially will be composed of a multidisciplinary team um both of course you know scientists but also your you know your lawyers as well and we currently have an RAI executive council that kind of is your strategic I guess um governance body that's overseeing RA RAI in the company and across all of our portfolio companies. We then have a number of of other kind of subsets including an RAI review committee and that committee is kind of touching upon some of the things that Magda said. They're focused at looking at the products um you know testing them. We do a lot of red tape testing as well you know as Danny had alluded to to ensure that you know if we can fix a problem now before it goes out to market then you're you're kind of saving on you know potentially millions of litigation in the future.

Um we also have a model repository um which allows us to kind of store all the models but also um keep a record of all of the risks that have been identified and and and that allows you then to also identify potentially patterns and deal with issues as as they arrive. So there's a number of things that that we're doing. We also have our AI ambassadors and including you know our legal team and myself. um some of us having you know to upskill a lot in AI and being sent to I've done a course in AI MIT I've got team members going to the Oxford Business School. So it's actually becoming an integral part now of our day-to-day jobs.

And then of course you looking just on the purely on the legal side. Um it's a little bit like the approach to the Exante regulations you know we don't know yet how liability will be aortioned and who will be blamed but contractually if we can try to kind of reduce the risk um that's what we are focused on right now. So, you know, as developers, for example, some typical some typical things we'd be looking for would be, you know, um maybe some undertakings on the deployer or user as to how they're using the AI. Um and there'll be restrictions on how they use that and if they breach that, then that will probably be carved out of the indemnity. Um we'll look at audit rights as well, so we can monitor that.

Um on the other hand if we're using or deploying AI we'll want representations and warranties um from the other counterparty particularly around the use of training training data um IP infringement. So I think that's predominantly what we're kind of focused on at the moment.

Thank you. That's really interesting from I mean I'm I'm a non AI guy but I just found all that really interesting everything that you sort of have to think about in terms of your your sort of global business. Um I'm just conscious of time. Um closing remarks uh just from each of the panelists. What is it you think uh that will change most dramatically in the next 5 years in terms of the disputes landscape and the use of AI?

Let's start with you man.

I think one one word we haven't mentioned but I think a lot of it comes back down to this is trust. We've just got to be able to trust AI and the more we trust it, hopefully the fewer disputes we'll have and the more we'll be able to embrace its benefits. And I think all of the comments that we've had on stage, I think go to the trust point. I mean, just to give you a quick example, I flew here on an airplane. I have no idea how that airplane stays in the air safely, but I still trust it.

And I I wonder why that is. And I think a lot of it is because of good governance. Yes, there's regulation, but there's also responsible practice and good governance. There's a safety card. There's a safety demonstration.

I know the pilots are trained. I see someone checking, inspecting the plane before we go. And I think the whole responsible AI movement, which is unbelievably important, is all about creating trust in AI. And that's going to be the currency and I think really important for the future and avoiding disputes.

Denny trust.

I'll give a negative and a positive. Negative. Evidence. Deep fakes are going to make litigations so much harder because you can now fabricate evidence with almost no cost. That's the bad.

The good, access to justice. Litigation. I've been a litigator my whole career. I've had cases last 10 years. That's too long.

Justice takes time. It's too slow. It's too complicated. AI is going to give more people access to efficient justice more quickly. Magna,

I think just on a couple of points that were made in terms of the ubiquitous nature that we're going to see AI, you know, if AI does become electricity, of course, we'll see an increase in disputes if it's everywhere. We're all using it all. How many times has everyone looked at their iPhone during the time we've been talking? You know, we're all using it all of the times. Hopefully not much.

Yeah, you were listening very intently. Uh so so of course there's going to be be more disputes and I think as time goes on with the trust element that you just mentioning with people uh evidencing what they're doing, why they're doing it, how they're doing it, um then I think we'll be able to um have a less latigious uh view of those disputes.

Donn word to you.

Yeah. So with um disputes increasing I think over the next five years you know what I'm most interested in seeing is um if we can actually end up maybe not exactly with a market position but right now there's a lot of uncertainty right and lots of jurisdictions approaching AI in a different way and different outcomes in cases and I think you know for common law to develop you need in a way more more cases more disputes and you need more facts so that you can end up in a situation where perhaps we can create a little bit more certainty um and so that that's what I'd like to see because with c with certainty perhaps over time we can also then in turn reduce the litigation.

Great da Danny Mesh thank you very much. I've really enjoyed it um and learned a lot. Thank you. Thank you. [applause]