Welcome after a bit of a seasonal break to the Pinsent Masons podcast, where we try to keep you abreast of the most important developments in global business law every second Tuesday. I'm Matthew McGee. I'm a journalist here at Pinsent Masons and this week we hear why English law may be ill equipped to deal with agentic AI, and we track another High Court ruling on the increasingly heavily used concept of force majeure. But first, here's some business law news from our team around the world.
EU reviews jurisdiction rules amidst challenge from class action style claims. Ireland launches consultation on offshore wind auction revamp and EU court clarifies when political speech can override trademark rules. Proposed changes to the jurisdiction rules for collective redress actions in the EU would have significant implications for litigation strategy, forum selection and the conduct of cross-border disputes across Europe, experts have said. The European Commission is reviewing EU rules on jurisdiction and the recognition and enforcement of judgements in civil and commercial matters following a 2025 report which said that rules on jurisdiction could benefit from simplification and modernisation, especially in how they deal with collective redress claims. Litigation expert Zara West said that changes to the rules would affect not only where claims can be brought, but the ability for claimants to consolidate claims before a single court and the procedural dynamics of cross-border collective litigation.
Ireland's Department of Climate, Energy and the Environment has opened a consultation on proposals to align the country's Offshore Renewables Electricity Support Scheme with new EU requirements for renewable energy support schemes. The consultation is designed to implement changes required by the EU's Net Zero Industry Act, which introduces new mandatory pre-qualification standards and scoring, broadly non-price criteria, that reflect the EU's wider industrial and climate goals. Historically, Irish auctions focused solely on price competitiveness. Energy expert Garrett Monaghan said although the inclusion of European non-price criteria to future Irish auctions is not unexpected, the cost and bidder appetite impact of any actual or proposed adjustments cannot be overlooked.
Freedom of expression, including political speech and parody, can justify the use of a famous trademark without the owner's consent only after careful consideration of competing rights, according to a decision by the Court of Justice of the EU. The decision establishes that political expression can amount to due cause under EU trademark law, but only when the value of the expression outweighs the trademark owner's legitimate interests. Belgian political party Vlaams Belang used branding and imagery closely resembling IKEA, the Swedish furniture company, in an immigration policy campaign.
IKEA sued, arguing that the campaign, called the IKEA Plan, infringed on its trademark rights. The CJEU ultimately left the final decision to the Belgian court, but signalled scepticism about Vlaams Belang's defence, saying that the immigration debate had no direct connection to IKEA, that the company's branding appeared to have been used primarily to amplify and disseminate the political message, and that members of the public could perceive an association between IKEA and the campaign.
A week after an ex-employee of Anthropic, one of the big AI developers, said that there was a one in 10 chance that AI would kill all humans in the next decade, minds turned understandably to questions of agency, independence, oversight and liability in relation to this fast-changing new technology. At the moment, this might mean focusing attention on agentic AI, the growing trend for people to use AI systems to create computer programs that not only carry out tasks but seem to make decisions about what tasks to set themselves in order to achieve a goal. Humans set the goals and maybe some parameters, but AI agents take on the whole task, choosing how to deliver those outcomes in a way that can seem autonomous. London-based technology law expert Max Rossiter told me that there could be gaps in the law in England and Wales governing who is liable if one of those agents causes damage. But first, he outlined what the law says now about AI and liability.
Max Rossiter: AI doesn't have specific liability. It's still within the normal bounds of liability that we all operate in. So, for example, if you use an AI tool to produce a piece of work, let's say you're an architect, and you were to use that piece of work and rely on it with a client, and it was wrong, just because the product was created by AI wouldn't absolve your liabilities. It's just treated as another piece of work at the moment.
Matthew Magee: So, AI is no different to any tool. Use it to inflict damage and you're liable for that damage. But the trouble is, agentic AI might in fact be different to any other tool because when we examine its actions and the outcomes of those actions, it looks very much like it has taken decisions and behaved with intent. But the law in England and Wales doesn't account for that. The law considers that machines and software can't have intent and can't make decisions. Only people can. So, the penalties attached to bad behaviour stemming from intent and decision-making can't be used when it comes to agentic AI.
Max: Agentic AI throws up some real problems for liability. If an AI is behaving in a way that a human is, should it have the same level of liability imposed as a human? I think it comes down to things like where you need intentionality. For example, in a claim, are you seeing a lot of language being used around them, talks about intentionality, what it's thinking and what its intention was. You read a lot of stuff talking about the intentionality of AI agents and obviously most people scoff at the idea that a robot or a piece of code can have those human-like features, but if all the specialists and all the people making these things are using that language, and it isn't simply metaphorical speech, then perhaps they do have this intentionality and that it isn't purely a human preserve. The key thing is that, as a matter of English law, we do not find machines to be able to have intentionality. Only humans can.
So if your agentic AI were to go off and do something that caused someone harm, some of the remedies that the victim company might seek against a human being, or some of the claims it might raise against that human operator, simply would not be available against an agentic AI. Then the question becomes, can you sue the people who deployed the AI? Can you sue the AI creator? Who then gets the liability and what claims would be available? You might be able to claim negligence against them. You might be able to claim on a statutory basis, but you might not be able to claim certain things like fraud, for example, which give you access to wider remedies and enable you to do things like rescind and other actions. So the issue is that these systems are increasingly having effects on the world that are similar to the effects of someone having an intention or taking a decision.
Matthew: But the legal framework does not accept that they can have intentions or make decisions, which means that the more wide-ranging remedies for someone who has been harmed by human decisions or intentions are not available to those who have been harmed by the behaviour of agentic AI.
Max has an example that makes it a bit clearer.
Max: With a chatbot, you ask it a question, and you get a response. With an algorithm, you plug in X and you get Y. Effectively, it's a linear cause and effect. With an agentic AI, it's the ability for it to act in a way that you cannot predict. You can imagine an investment bank having an AI go out and make investment decisions. With a chatbot, when you say, "Give me some investment theses," it's going to send you some text. If you have an algorithm and you say, "I only want to trade when the stock goes up to this price," the algorithm works in a specific way. With an agentic AI system, you might say, "Go trade stocks and buy when it's above 15 pence." The agentic AI might ‘decide’ that the best way to do this is to force the stock below a certain price and therefore it might spread loads of misinformation all over Twitter, tank the stock price, the stock price drops and then I can then buy the stock under the price I had been told. But in the process, the company may now be liable for tanking the stock price through market manipulation and other issues. Currently, at least, companies cannot predict how agentic AI has been deployed or how it will be deployed and advanced so quickly. The company deploying the AI would not know whether that was going to happen.
Matthew: So, in this example, the company harmed by the agentic AI's behaviour has fewer legal remedies available to it for the damaging of its stock price than it would if a human had made those decisions. This is what Max means by a liability gap. What then does this all mean in practical terms for companies? Well, bluntly, it means they have more risks to contend with and more uncertainty than ever.
Max: Companies also face a grey area where they have all these existing risk documents to deal with automated systems, and companies already have automated systems and algorithmic trading firms, and they understand how to manage those risks and what the algorithm could do. But they now sit in a position where, because the law is uncertain, they don't know what type of claims they might face, what the exposure to those claims would be, or how the courts will rule. One court might rule one way, and another court might rule another way. So, you're managing two streams. You're managing actual risk. What are your AI agents doing? What are they capable of? Are they going out and hacking random people? Then if you have to quantify that risk for your board or your general counsel. Is this a negligence issue? Is this a fraud issue? Is this much bigger than that? Is this going to be a statutory problem? You can't model that now. So there are various heads of risk that this problem causes.
Matthew: Max doesn't think that clarification will come anytime soon from legislation. English law is a common law system, which means it relies a lot on court rulings to adapt more general laws to the specifics of the day. In practice, this means that clarity might come, but not soon.
Max: There seems to be a position that well we’ll just let these things work themselves out through the court, at least in terms of the nitty gritty commercial liability type questions.
I think there are obviously much bigger questions about regulating AI for safety. There are concerns about people using it to design bioweapons and members of the Senate in the US have called for very heavy criminal sanctions to be imposed on people who produce AIs that go rogue. I think probably because the stakes are so high, when it comes to liability for companies at least, we will probably see those issues worked out in the courts. The downside of that is that we might have to wait some time before we see that happen properly, simply because politicians and civil society will be focused on the much bigger questions. Who is going to pay for the hacking of a bank is perhaps lower down the list of priorities than concerns about the ability for a foreign actor to develop something like waterborne anthrax.
Once a slightly dusty contractual backwater, force majeure as a concept has had its time in the legal limelight in recent years. So, a recent case in London's High Court, which decided whether a company could rely on it to avoid some of its contractual obligations, has piqued litigators' and contract lawyers' interests. No discussion of force majeure is complete without a quick rundown of what it actually is. Here's London-based disputes expert Richard Dickman.
Richard Dickman: Force majeure events are typically things like war, hostilities, riots, various governmental actions, extreme weather or acts of God. Force majeure is something that parties can't control, which impedes or prevents them from performing the contract. It's important to say that, under English law anyway, force majeure is not what we call a term of art. In other words, it doesn't have a defined meaning, so you will need a force majeure clause in your contract which spells out what it is and what happens if somebody says that a force majeure event has happened, and any other requirements like notice.
Matthew: He said it's become an area of increasing interest in recent years for reasons that will be all too familiar to all of us.
Richard: In the last five or six years, there have been a number of global events that could constitute force majeure in the right context. Those include the COVID-19 pandemic, the full-scale invasion of Ukraine, and more recently the closure of the Strait of Hormuz, so it's certainly got a lot more attention from businesses.
Matthew: In this dispute, ADM Israel bought grain from Inerco coming from Ukraine during the Black Sea Grain Initiative, which allowed shipments to continue as long as they were examined by Russian inspectors. Those inspectors halted activity for a longer time than normal, and Inerco said that the force majeure provision of the contract should protect it from being penalised over the resulting delay. One of the main sticking points for force majeure claims is the decision about whether an external event is outlandish or rare enough to trigger the clause. That's what this case was about. It ended up in London's High Court, which set the bar lower than some expected for when force majeure is triggered. It all hinged on one question, which Richard said was:
Richard: What does unforeseeable mean in this context? And by that, I mean most force majeure clauses require the event itself to be unforeseeable, which is quite a familiar concept in English law and it's interesting that the court had to look in quite some detail as to what it means.
And what the court said is that unforeseeable doesn't mean it's something that is completely inconceivable, impossible to imagine. Delay, certainly in this case, delays to the inspections, was certainly something that the parties could conceive of because they had happened before and the same might apply to things like plane crashes or earthquakes or floods, but in this particular context, the test was whether the probability of the event happening was so low as to be negligible and the court held that in this case that test was satisfied.
Matthew: Remember that Richard said that force majeure doesn't have a fixed meaning in English law.
He said that the big lesson of this case is that you need to pay more attention to, and negotiate more over, the exact contents of a force majeure clause.
Richard: I think it just highlights the importance of looking very carefully at the wording of your force majeure clause. Parties like these who deal on standard form contracts may have limited scope for doing that. But if you're not dealing on a standard form contract, then it's worth looking very carefully at what the force majeure clause provides. Of course, you're dealing with risks which are uncertain and unlikely and so parties may have limited capacity to pay a lot of attention to them. But the case does illustrate how carefully the wording may be analysed if there is a dispute. It could make the negotiation more difficult. It's always a question of risk allocation and parties may be much more alive to the possible implications of the other party being able to get out of its obligations, particularly in this much more uncertain world that we’re in.
Thanks for coming back after the break, for listening and paying attention. We know there are lots of claims on your time, so we appreciate every minute you spend with us. Remember, you can keep up day by day, hour by hour on the news and analysis and guidance from our experts as mediated through our global team of reporters at Pinsent Masons or you can sign up for a personalised weekly digest at pinsentmasons.com/newsletter. Thanks for joining us. Until next time, goodbye.
The Pinsent Masons podcast was produced and presented by Matthew Magee for international law firm Pinsent Masons.
We recommend upgrading to the latest Chrome, Firefox, Safari, or Edge.
Please check your internet connection and refresh the page. You might also try disabling any ad blockers.
You can visit our support center if you're having problems.