Hello and welcome back 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 Magee. I'm a journalist here at Pinsent Masons. And this week we welcome some movement finally on public policy governing the tension between AI system development and copyright law. And we learn about the biggest overhaul to a vital UK financial consumer protection in decades. But first, here's some business law news from around the world:

China tax rebate changes could affect solar and battery exports
UK competition regulator examines energy price hikes during Middle East conflict and
Ireland looks to raise turnover thresholds for mergers

Changes to China's export tax rebate system will materially affect global supply chains, including for solar panels, an expert has said. China will revoke export tax rebates for photovoltaic and battery products. This will significantly change the position of foreign companies sourcing renewable energy components and industrial materials from Asia. Chinese manufacturers originally received 13% VAT rebates on export products and currently receive a nine percent VAT rebate. With the elimination of these rebates, they will no longer be able to claim that this will result in an increase in the cost of exporting products. Energy law expert Jurg van Dyk said that manufacturers and buyers would have to collaborate and that project owners, employers and that contractors will also need to carefully assess how the reductions to export tax rebates may affect their existing contractual arrangements.

UK competition regulator the CMA is scrutinising recent price surges in heating oil and road fuel after the Middle East conflict brought major disruption to global energy supplies. The CMA is using its wide-ranging competition law and consumer protection powers, which were strengthened by a new law last year, to gather information from heating oil distributors and intermediaries to investigate concerns that companies are profiteering from the crisis at the expense of consumers. There have been reports that some heating oil suppliers in the UK abruptly cancelled pre-existing orders only to offer new contracts at significantly higher prices. And the automatic top-up deliveries, which are triggered by low tank levels, were suddenly being charged at steep mark-ups. The regulator said it's written to a number of firms, both direct suppliers and intermediaries, to seek further information about their practices.

Plans by the Irish government to raise turnover thresholds for merger notifications could result in more below-threshold deals being called in for scrutiny, an expert has said. The Irish government has launched a public consultation on the proposal to increase the turnover thresholds that would trigger mandatory reporting. The increase for individual turnover thresholds is from €10 million to €15 million and an aggregated threshold from €60 million to €100 million. The new proposal is expected to save Irish firms around €1.8 million a year in reduced costs and fees, with Irish Enterprise Minister Peter Burke claiming the proposed increase would allow the Competition and Consumer Protection Commission to place greater focus on mergers with potential to have a distortive effect on competition. However, mergers expert Paul Williams said that against the backdrop of the regulator's call-in power, which enables the authority to require notification of transactions that fall below the thresholds, the proposed increase in those thresholds may in practice lead to more deals being called in for review, creating additional challenges for companies.

We've talked a number of times on this podcast about the tension governments face on the training of AI systems on copyrighted material. On the one hand, they want to attract AI investment by making the development of AI systems easy, which means allowing the use of copyrighted material to train them. But they also want to support the creative industries and their income streams, which means not allowing the use of copyrighted material in that training. They're in a bit of a bind and an air of paralysis has lingered over the policy area for a while. But all of a sudden movement. The EU, US and UK have all shifted their position in recent days and some clarity is emerging. So we turned to our resident AI training and copyright tension expert, London-based Gill Dennis, to fill us in on what's happened and what it all means. So what's changed? The UK has rolled back on a plan to extend an exception to copyright law for AI training, says Gill.

Gill Dennis: In the UK, we had the government publish a mandatory report that it was required to do by statute, really outlining its future policy on copyright and AI. So we were holding on, hoping this was going to give us a clear way forward on where we're going to be on the use of copyright works to protect AI in the future. And the policy paper was a bit of a disappointment from that perspective because effectively what the UK said is that we are going to sit tight and wait and see. I'm not going to do a great deal at the moment. So the government explicitly abandoned its plan to introduce a broader exception, a text and data mining exception for the use of copyright works to train AI. The government's plan was to introduce a wider exception that would allow AI developers to train their tools quite freely on copyright work, provided that rights holders were given the option to opt out. So right holders were going to be allowed to have some control over their works. But the government has, many say, retreated from that. At the moment we don't have a clear way forward in the UK. It's a wait and see. They're looking to gather more evidence on the possible effects of changes to copyright law. They're also really waiting to see what's happening in other jurisdictions as well.

Matthew Magee: The US has been bolder and has indicated a clear direction of travel and it's towards the AI developers.

Gill: In the US, we've had a much more clear steer on where we're going on this issue. Just in the last week or so, the US government has issued a national policy paper on AI and copyright. The paper makes it very clear that in the US government's view, training AI tools on copyright works will not be copyright infringement. That's the kind of presumption that's been introduced. However, the paper does go on to say that it recognises that there are other views and that there's still a debate around this. So they have left it open for rights holders to be able to take a contrary argument to the US courts. So although we begin with this initial presumption that it's all OK, rights holders can go before a court to say actually it isn't OK. This is not fair use of my copyright work. And that's the test. Is the use fair use? And ultimately it will be down to judges to decide that on a case-by-case basis.

Matthew: Things are a bit more complicated in the European Union, where policy, legislating and government is split between the member nations, the Parliament and the executive arm, the European Commission. The Parliament has come out fighting for the creative industry more strongly than anybody would have predicted, says Gill.

Gill: In the EU, we've had actually quite a tangential shift. The EU has been very clear in recent times through legislation that using copyright works to protect AI is lawful under a TDM exception, a text and data mining exception. The Digital Single Market Act says that AI developers absolutely can use copyright works to train their tools, but rights holders have the option to opt out using machine-readable technology. So that seemed to be quite clear. But then slightly left field, on the 10th of March, the European Parliament passed a resolution saying look, this just isn't working in the context of AI. The copyright regime doesn't sit well and we need to review this. So the European Parliament made some quite arguably controversial suggestions for a shift in the legal position much more in favour of rights holders. Making transparency obligations on AI developers much more stringent. The European Parliament's policy paper has really been seen as quite controversial. But there it sits, with a suggestion that the EU really needs to review its regime and potentially revise it.

Matthew: It's not an easy picture to grasp with everybody moving in different directions, but Gill says that things really are clarifying and that we can see some daylight between the policy positions of the EU, the US and the UK, but also some similarities.

Gill: The debate in the three jurisdictions we've been talking about has really shifted from that debate around is the use of copyright works infringement at all, which I appreciate is the question that underlies all of this. But we seem to have come to a place, either expressly or implicitly, in all three jurisdictions that potentially it is copyright infringement to use copyright works to train AI. That was a question we had absolutely no answer to perhaps even 12 months or even six months ago. There is definitely some acknowledgement that that's an issue in all three jurisdictions. But I think where we have a shift is in the US. We're seeing a shift towards actually it's going to be OK to use copyright works unless you show us to the contrary. The EU now seems to be taking a position of actually it is copyright infringement, but we've introduced a broad TDM exception to protect both AI developers who need the works and rights holders who need the income. Perhaps now, with some shift shown by the European Parliament in terms of more stringent obligations on AI developers and more protection of rights holders, there could be a shift there. And I think the UK is somewhere in the middle, saying actually we've not sided with either. We recognise that we want to protect either. We're not shifting legislation at the moment to move in either specific direction. So there is some division between those jurisdictions.

Matthew: So, might we see some countries emerge as hubs through the training of AI systems which are then used everywhere, even in countries which have said that the training violates copyright law? Certainly in terms of the UK position, even that isn't so simple, says Gill.

Gill: So I think what we may see in the short term is training in the US where there's a presumption that it's OK, or so it appears, and relating that to the UK. I think what could happen then is that those models are deployed in the UK rather than trained here. What is unclear at the moment in the UK is the extent to which that deployment will be lawful. So we have the appeal in the Getty Images case. Getty Images and Stability AI are going to the Court of Appeal. The hearing is listed for later this year, where the question will be the legality of importation of tools that have been trained on copyright works elsewhere. The current legal position here, based on the High Court decision in Getty, is that it's lawful to import and deploy those tools trained elsewhere, if the machine, the tool, the model itself doesn't store or reproduce those copyright works. If it does, then the importation and deployment may not be permitted. So that's something that needs to be worked out in the UK.

In relation to the EU, I would suggest that training can still take place there. There is still that TDM exception, unless some of the European Parliament's policies or suggested policies are implemented, which may mean that EU development takes place more outside that jurisdiction going forward, but probably less of a shift from the EU than in relation to the UK.

If you offer financial products to consumers, you can expect your activity to be heavily scrutinised and quite right too. There's a network of checks and balances that in the UK is about to change quite dramatically. The Financial Ombudsman Service is there to receive, investigate and act on consumer complaints about financial firms. It's one part of the consumer safety net. But the financial industry has been complaining for some time that the boundaries between its work and that of the regulator, the Financial Conduct Authority or FCA, are confusing and blurred. The UK government now agrees. It thinks the Ombudsman has been behaving too much like a regulator and has introduced sweeping changes to address that problem. Before we get to the changes, I asked London-based financial regulation expert and former Ombudsman himself, Anthony Harrison, where the Ombudsman fits into the wider financial regulation picture.

Anthony Harrison: The Financial Ombudsman is basically an independent body that's set up to resolve disputes between consumers and financial services firms without the need to go to court. It's really there to be a more accessible, informal process than a court process. It's free for consumers and they essentially look at financial service complaints and whether the firm complained about acted fairly and reasonably. A regulator and an ombudsman serve different purposes. A regulator, and in this case and in the context of financial services, the main regulator is the Financial Conduct Authority and that is a body which, amongst other things, sets various rules of conduct and compliance that financial services firms have to follow.

So they set the rules and they also act as a kind of watchdog, making sure that the financial services market and the firms that operate in that market behave appropriately and adhere to those rules. In contrast, the ombudsman is not a rule setter. What its purpose is, is really to be the kind of interpreter of those rules, and it looks not so much at the macro picture, but at individual complaints.

Matthew: So what was the problem? How was the Ombudsman stepping into the FCA's territory?

Anthony: An ombudsman, when they issue a decision, a final decision, it is legally binding on that regulated firm that then has to follow what the ombudsman has said. The challenge is regulated firms, when they receive an ombudsman decision, not only do they have to apply it to the actual case, but they also have to consider, well, what is the read across of that decision. They have to consider applying the Ombudsman's decision to potentially a huge book of similar cases and so that one individual case can have a real sort of ripple effect in terms of a firm's compliance and indeed how a market might have to follow it. So if you have a number of firms that are faced with a systemic issue that lots of consumers are complaining about, then actually an ombudsman's decision can have considerable market impact. And so it starts to move into that macro space that I was talking about. It starts to have that wider market impact and it is not so much contained within the boundaries of an individual complaint. You have the ombudsman effectively shaping the standards that firms have to meet, even though really it is the FCA that has the power to set the rules. And so you have this sort of blurring of those two roles.

Matthew: The UK government has set out what it wants to do in a lot of detail, but Anthony said there are broadly four main changes.

Anthony: First and foremost, a rebalancing of the fair and reasonable test. Has that firm acted fairly and reasonably. Whereas before the Ombudsman has applied that test by reference not only to FCA rules but to case law, to industry guidance. The rules are now saying, well, if a firm has complied with the relevant FCA rules, that should carry real weight. So what it is doing is it is narrowing the ombudsman's discretion and it is anchoring it back to those FCA rules.
A couple of other big things that have come through or are being proposed is a formal referral mechanism. So if the FOS thinks a rule is unclear or has those wider market implications, it can formally refer that complaint to the FCA so that the FCA can look at it in terms of its rule interpretation. Not to decide the complaint, but to help give a steer on the rule interpretation where there is ambiguity and a concern that actually if the Ombudsman interprets the rule in a particular way, it could have wider ramifications.
The third thing is a ten year absolute long stop. So that is essentially a time limit, whereas before there has been a bit more wriggle room around time limits. It is called the six and three rule, but basically means either your case would be time barred if it was about something that occurred more than six years ago, or where you brought your complaint more than three years after you realised you had something to complain about. If you apply that ten year absolute long stop, a firm should not face complaints stretching indefinitely into the past anymore.
Finally, the fourth thing that has sort of come out of those proposals is a kind of triage phase where complaints can be filtered out more easily at the front end that are not necessarily ready for investigation. And that stops situations where claims management companies might have harvested thousands and thousands of complaints on behalf of customers, threatened a firm that they would bring it to the ombudsman in the knowledge that by bringing it to the ombudsman, each case attracts a case fee. And it varies. I think it is six hundred and fifty or seven hundred and fifty pounds per case. Well, if you have accumulated thousands of cases, you have a real sort of bargaining leverage. And so this is sort of designed to try and rationalise that and make it such that it does not necessarily follow that just because something is referred to the ombudsman, it will automatically incur a fee, because there is better triaging upfront.

Matthew: The government will have to pass new laws to make these changes, so Anthony thinks they will not take effect for at least a year. But companies should prepare now for the changes ahead, he said, and maybe even take them into account in dealings with the ombudsman and regulators ahead of their coming into force.

Anthony: If rule compliance becomes central in terms of showing that you have complied with the FCA rules, it is about ensuring the quality of your audit trails, your MI, your rationale documents for why as a firm you have done certain things in a given case or situation or product, so that you can show and demonstrate the necessary compliance for the ombudsman. Some of it will be about mapping your legacy exposure and your complaint profiles, especially if there is a ten year time bar coming in. How does that affect your book? If you have got longer legacy cases, might they be caught by exceptions to that ten year long stop?

Again, thank you for listening and please do share this and other episodes that you have enjoyed with anyone you think it might be relevant to. Do review it if you feel like it. And remember, you do not have to wait for the podcast to hear what is happening. Our team of dedicated reporters are working all day, every day to bring you up to the minute reports from all around the world about what matters in business law news. You can find all that at pinsentmasons.com or you can subscribe to a personalised newsletter at pinsentmasons.com/newsletter. Until next time, thanks for listening and goodbye.

The Pinsent Masons Podcast was produced and presented by Matthew Magee for international professional services firm Pinsent Masons.