John Patten: Yeah, thank you both and delighted to be back. So, as I mentioned on our previous discussion, my practice involved advising on a number of digital regulations across the EU and the UK. Alongside the EU AI Act, UK AI advice has become a fast growing part of the practice. A lot of English law is of course originally derived from the EU, so there are plenty of parallels, but since Brexit the two are pulling apart more and more and so AI is no different in this respect.
Scott Caravello: Yeah, and so you know, last time when we were talking about the EU, the big headline was, you know, the EU AI Act, the big comprehensive risk based statute and the recent developments around it. And so, to your point about that, it seems like the UK story is a very different animal.
John Patten: Completely different. And so the single most important thing to understand about the UK from an AI perspective is that as of today, there is no AI Act equivalent and no dedicated AI regulator or regulation at all. Whereas Brussels built one comprehensive law and Washington has issued various executive orders, when US states have made some progress on the legislative front, the UK has deliberately chosen not to regulate AI as a single category. Instead, the framework rests on existing laws across IP, privacy, online safety, product liability, et cetera, each of which is enforced by the existing regulators or the courts.
Katherine Forrest: And is that, John, in your view, a deliberate strategic choice?
John Patten: Yeah, I think very much so. So a large part of the UK's pitch is that it's a bridge between European, US and other international markets. English law and the English courts are popular and predictable, and the UK has historically had strong ties to both the US and the EU. So, for many businesses, it's a natural conduit between two jurisdictions. That does result in a bit of a balancing act as a step towards one superpower's approach might mean a step away from the other.
Katherine Forrest: Wait, hold on. Before you go on, Scott, I have to just tell you that have we ever had anybody here speak in an English accent who uses the word “conduit” in that way? I just I could listen to John like just repeat that word conduit like twenty times. It would like it would be like just fun.
John Patten: I'm very happy to play my part.
Katherine Forrest: I've rendered you speechless.
Scott Caravello: Well, we do need to see if we can work what was it, long stop date? Was that was that the term we got hung up on last time? We gotta work that in there. But you know, before we get into the nitty-gritty of the regulation or, you know, the balancing approach, maybe to set the stage we can talk a little bit about the commercial landscape in the UK. Because, you know, I mean, here we talk about Silicon Valley all the time. But really, I mean, the UK is a genuine AI hub in its own right.
John Patten: Yeah, it really is. After all, DeepMind was born in the UK and pioneered many of the early steps in AI and machine learning globally through AlphaGo, AlphaZero and AlphaFold. Now DeepMind is instrumental to Google in their Gemini models. Then we've got Stability AI in image generation, Wave in self-driving, 11 Labs in voice and Humanoid in robotics, and they're all UK based. The investment story has only really accelerated since then. So we've just wrapped up London Tech Week with more than £6 billion of new AI investment and around 8,000 jobs announced, including AMD committing up to £2 billion for UK chip design, R&D and innovation, and Nebius deploying £1.7 billion in the video GPU clusters across three new UK sites.
Katherine Forrest: Well, you know, that's actually very interesting because you've got sort of these job announcements and we hear so often about job cuts, but 8,000 new jobs announced, that's a that's a pretty big number. And the UK government is putting in its own money, at least that seems to be the press reports. We were reading about how the government is putting in over a billion dollars for something called the AI hardware plan.
John Patten: Yeah, that's right. So that sum includes 400 million pounds for British made AI chips and 750 million for a national supercomputer, which is meant to be in place by 2030. That sits on top of the 500 million sovereign AI unit that went live in April, which invests in UK AI startups. There's also another example in the UK government's investment in quantum technology to support AI with their announcement in March this year for a plan to invest two billion pounds in quantum innovation. And this is already showing benefits with large quantum businesses like IonQ already setting up collaborations in the UK.
Katherine Forrest: You mentioned that there's no UK AI Act and you previewed it a bit before, but there is some regulation in the UK, so how is that being accomplished?
John Patten: So the answer is that other than kind of specific IP privacy, et cetera, laws that govern certain aspects of AI training, use and output, there is no central AI regulation. Instead, the government's focus has been on investment. And this has been via the AI opportunities action plan, which was adopted in January ‘25. And this is aimed to drive investment to position the UK as the best state partner to the companies building frontier AI. Interestingly, it's actually already shown tremendous steps with compute capacity up roughly tenfold and five AI growth zones that are set to generate 28.2 billion pounds of investment and more than 15,000 jobs.
Katherine Forrest: So again, 15,000 additional jobs on top of the 8,000 jobs. So really interesting stuff. And so what I'm also hearing is that the UK is making some serious investment in the AI infrastructure that we've been talking about on prior episodes and also pushing adoption. So what's the outlook on that front as we move forward in the rest of 2026?
John Patten: So from a legal perspective, the way that this is then flowing through is that there's a more conceptual-based regulating for growth bill that's going through the legislature, which is aimed at promoting innovation in regulatory sandboxes. And there's also a live private members bill, as we would call it, called the Artificial Intelligence Regulation Bill, which in theory could take a more centralized approach like the EU AI Act. This proposes a central AI authority with principles based statutory obligations, e.g. safety, transparency, accountability. That second bill had a parliamentary debate in early June, but it doesn't have our government's backing, a private member's bill is generally less likely to become law. But it is a useful marker that there's an ongoing debate about whether the UK should move towards a statutory model.
Katherine Forrest: And let's talk about copyright for a moment. In the United States, I mean everybody has, you know, been talking about the large number of copyright cases and there's just a host of them now. Many of them are but not all are have been centralized in the Southern District of New York, in the federal court here and they're looking at the question, among other things. But the central question is whether or not the use of data that might have proprietary interests, as, for instance, copyright is, can constitute when it's used for training, fair use. So, a use that could occur without permission. And so this is going on in the courts right now and closely watched in the UK as well. And you've actually had some significant decisions make international headlines. So let's start with the legislative side of copyright since there's been since there's been activity in the UK on that and then cover, you know, where you are on the sort of the judicial decision front.
John Patten: Yeah, absolutely. So maybe on the legislative side, where we should start, it's that this has been a very closely debated area. And the government recently spent nine months in a consultation considering what update should be made to the legislation. And the upshot is status quo. Nothing changes, which is not unusual for our government. So ,no new legislation, no new regulator for now.
Scott Caravello: But that's a change from where the government started, isn't it?
John Patten: It is. So the government's original preferred option had been one to align to the EU, which is a broad exception to copyright infringement, allowing commercial text and data mining subject to the rights holders being able to opt out. But that was opposed by most respondents to the government's consultation with only 3% backing that government preference.
Scott Caravello: So then with that in mind, where does that leave the creative sector and then what comes next?
John Patten: So the creative sector has made some strong statements. Elton John has publicly described unlicensed training as theft and a silent album was released under the names of more than a thousand artists in protest of using trained data without express permission. But rather than legislate now, the government has set out where it will focus next instead. So that's digital replicas and personality rights with a consultation launching this summer, the labeling of AI generated content with a task force due to report in the autumn, and a proposed creative content exchange to make licensing easier.
Katherine Forrest: So that's the legislative side, but how about the judicial decisions? Where are the courts these days?
John Patten: Okay, so there are three court cases to know about. The first is Getty Images and Stability. This is the big training data copyright case that you've discussed before on this podcast. At first instance in late 2025, Stability essentially won. Getty had dropped its primary copyright claim on the basis that there was no evidence of training taking place in the UK and lost its secondary copyright infringement claim on the basis that an imported AI model is not deemed to be an infringing copy under English law, really only succeeding on a narrow trademark infringement claim due to certain watermarking. In December, the judge granted Getty permission to appeal. And now the narrow question in the appeal court is what counts as an infringing copy? I.e. given that a model was built using infringing copies of training data, is the model itself an infringing copy, even though it doesn't still store those works? The High Court rejected that, holding that given a model as imported did not still hold the trained data, it is not an infringing copy. But commentators have flagged it's a novel and important point of law with potential ramifications for AI models and software more generally. That appeal is expected to be heard later this year with a focus on whether things like overfitting or similar memorization techniques will count and bring that model within the realm of an infringement claim.
Scott Caravello: And so then moving on to the second case, that's a patent case and that caused a a s significant shift in the in the law, right?
John Patten: Yes, exactly. This is the emotional perception case. So some quick background. The UK patent law says you can't patent a program for a computer as such. And the question, therefore, was whether an artificial neural network required for an AI model falls within that exclusion. The lower courts had gone back and forth on this. But in February, the UK Supreme Court held that although an artificial network is a computer program, it isn't automatically excluded from patentability, which is what the lower courts had decided. In getting there, the Supreme Court aligned the UK test for software patentability to the more permissive test used by the European Patent Office, just noting that that's not an EU institution, which is why the UK is still a member. So for anyone patenting AI inventions in the UK, this is a meaningful opening.
Katherine Forrest: Yeah, and the third case is the one about whether AI can be an inventor, which is you know, that's been an issue that's been contested in the United States with the position of the courts in the U.S. patent and trademark office being that inventors must be human, though there's been some guidance now from the U you know, from the US PTO that says that AI can assist, but really you've always gotta have what I'm gonna call a human centrally and critically in the loop.
John Patten: Yeah, exactly. So on the UK side, we've got the same line of cases involving Stephen Thaler that you have in the US, who is the individual behind the landmark AI inventorship case in the US. The settled position is that UK patent law requires an inventor to be a natural person. So an AI model cannot be named as the inventor. That hasn't changed. And this case now simply sits alongside emotional perception as the two bookends of the UK AI and patent spectrum.
Scott Caravello: Awesome. So with all of that said, why don't we finally just sort of turn to online safety and I think if you wouldn't mind, maybe it would be helpful to put the UK, EU and US approaches side by side.
John Patten: Yeah, absolutely. And it's an area that's moving fast. So like many jurisdictions, the UK has gone hard at non-consensual intimate imagery across both nudifier apps and sexual deepfakes. Since February this year, it's been a criminal offense in England not to share, not just to share, sorry, but also to create and even to request the creation of a sexually explicit deepfake without consent, even if it's never shared. The government is also moving to criminalize supplying the nudifier tools themselves, and sharing is already an offence under our existing Online Safety Act.
Scott Caravello: And then I maybe you can shed some light 'cause I had also seen what what seemed like a a very significant announcement about children's access to social media in the UK.
John Patten: Yeah, so on the 15th of June, the government said that from spring 2027, it will ban under-16s from social media, which is a situation we've seen in Australia and other jurisdictions. Romantic companion chatbots will also have to enforce an 18 plus age limit and intimate functions will be restricted for under-18s. Ofcom, who is our regulator for the communications industries, has also been told to fast track its work on age assurance. So we have the UK criminalizing creation of this content, leaning on the Online Safety Act and Data Protection Law. The EU AI Act, then which is what we discussed last time, goes at it differently, through an outright prohibition on the Nudifier apps and on general purpose image and video models put on the market without safeguards. And as we discussed last time, that bites from December this year under the digital omnibus.
Katherine Forrest: Hm. You know, that's that's really interesting, particularly the entire social media ban that takes effect in the spring of twenty twenty seven and all of the other pieces. But in you know, in the US we have the Take It Down Act, which focuses on the publication and platform takedowns of quote, non-consensual intimate in imagery, end quote, rather than its creation.
John Patten: Yeah, so I guess it's the same harm that is being looked at across both the UK, EU and US, but it's just being dressed in three different ways, which is exactly the type of fragmentation our multinational clients are now having to navigate.
Katherine Forrest: Right. I mean that's why, you know, when we talk about different AI regulations, getting the perspective of lawyers from different jurisdictions, John, just like yourself, are really necessary 'cause it it's critical to pull it all together because there are different threads in different places affecting different areas of the law.
John Patten: So, I guess if I were to summarize then where we are with the UK, so the overall picture is the UK is positioning itself as pro innovation and sector led with no big EU style AI statute and instead more and more energy is going into investment, sandboxes and getting regulators to lean into growth. The UK's bet is to be the conduit, well there you go, saying it again, between the EU's more prescriptive comprehensive regime and the more market led approach in the US. For our clients, the practical question is the familiar one for multinationals, not just in AI, but in other sectors. Do you localize compliance to each regime or generalize to the strictest one you're caught by?
Scott Caravello: And so then what's on the horizon to watch?
John Patten: So I the biggest thing is obviously the Getty appeal, which is expected to be heard late this year with a ruling in early 2027, as well as the government's consultation on digital replicas and an interim report on AI content labeling. These are both areas where English law could do with a bit more certainty, but they're also pretty contentious and will no doubt shape how the UK's AI market develops.
Katherine Forrest: Well, John, this has been fantastic. It's sort of a a whole capture of the UK picture at a high level all at once. So, thanks again for joining us. We're not gonna let you go away forever because we have to see how that diffuser works and check in on your use of the juice bar, but we've also got some of these developments that we're gonna wanna sort of circle back to.
John Patten: Thanks, Katherine. I'll send you a diffuser in the post and it's very nice for you to have me again. Really great to speak.
Katherine Forrest: All right, all right. Signing off now, I'm Katherine Forrest.
Scott Caravello: And I'm Scott Caravello. Don't forget to like and subscribe.
