Copyright and AI training: "An almost paradoxical situation indeed"

Contents

Training generative AI models is not text and data mining. This is the conclusion of the new open access study "Copyright and training of generative AI models – Technological and legal foundations" , which was commissioned by the Initiative Urheberrecht. This is important, as AI companies want to rely on this, among other things, to avoid having to pay creative people anything.

"We began our investigation with an absolutely open mind and were also aware that the topic had not yet been extensively examined from an interdisciplinary perspective," say study authors Tim W. Dornis and Sebastian Stober. Dornis is a legal scholar and professor at the Faculty of Law at the University of Hanover and completed his J.S.M. at Stanford, while Stober is Professor of Artificial Intelligence at Otto von Guericke University Magdeburg.

In addition to negating the so-called TDM barrier, the researchers also assume that training data and thus copyrighted works are reproduced within the AI models. "This is important for legal action against infringements in AI training and AI use," they say. In an interview with heise online, Dornis and Stober explain what this could mean in concrete terms.

heise online: Mr. Stober, Mr. Dornis, in connection with the question of whether AI training in its current form violates copyright law, there is always talk of so-called text and data mining, under which this supposedly falls. What is TDM and what should its possible uses be?

Sebastian Stober: Data mining involves the automated extraction of new information, patterns and findings from large collections of data. Text mining involves text collections. The possible uses are very diverse. The insights gained can form the basis for business models – when, for example, markets and customer behavior are analyzed. In politics, analyses of public opinion are regularly used, for which text mining is carried out in the background in social media. Data mining is also an important tool in science for practically all data-based questions.

However, as with any technology, the intended use does not always have to be positive for society. For example, the information obtained can also be used to manipulate people in a targeted manner, as was shown in the Cambridge Analytica scandal. Here, society is called upon to set clear boundaries as to which uses are undesirable.

Many AI companies now think that their AI training falls under the so-called TDM barrier because they seem to take this from the AI Act. In other words, they think it is covered by law. Yet the roots of the AI Act were established years before generative AI systems even existed.

Tim W. Dornis: I don't see it that way either. It cannot be inferred from the AI Act and the legislative materials, if interpreted correctly, that generative AI training should fall under the TDM barrier. The wording alone raises doubts. Above all, however, an in-depth examination of the principles and background would have been necessary – with a particular focus on copyright law.

However, this was still lacking at the time the AI Act was finalized. In other words: As our study shows, a thorough examination of AI technology is required. This has simply been neglected in the legal opinion-forming process to date. However, the debate should not stop at this point just for the sake of "convenience" and with the argument that anything else could jeopardize "European AI innovation". Nevertheless, this seems to be the trend in the current legal debate.

Why is AI training so much more than TDM? It's just computers scouring the internet and drawing conclusions.

Stober: On the one hand, we have to distinguish between data collection and training –, which is often carried out by different actors, and once a data collection has been created, it can be used by a wide variety of actors to train a wide variety of AI models.

On the other hand, the concept of AI training must be differentiated more clearly. In our study, we placed great emphasis on emphasizing that we are talking about the training of generative AI models. There are regulations for text and data mining that allow the collection of data for the training of AI models. However, we come to the conclusion in the report that the training of generative AI models in itself does not fall within the scope of text and data mining – because, among other things, no new insights are gained in the process. The trained models can only generate further data that is similar to the training data. This is therefore a completely different purpose. The exemption therefore does not apply here and that is the problem.

Published material can be provided with a reservation against TDM, as provided for by law. Wouldn't that be a simple solution?

Dornis: On paper, it seems to be a simple solution. However, the practical implementation is anything but effective. You don't even have to go so far as to ask how to deal with works that have already been published (e.g. books). Should these be subsequently provided with inserts anywhere in the world?

For digital publications, too, we will have to assume that once things are "online", they can hardly be subsequently provided with a complete – and, above all, for crawlers & co. understandable – with a reservation. Finally, the question remains (as always) as to whether the AI developers (and their crawlers etc.) adhere to this at all.

Can you explain how the AI industry could assume that the training in its current form represents a kind of fair use (which does not exist in Europe)? First do it, then apologize, as Mark Zuckerberg once recommended?

Dornis: From a legal perspective, it's easy to explain: The mindset in Silicon Valley has always been "don't ask for permission, ask for forgiveness later". It is less about the idea of the legality of one's own actions and more about the conviction that, in the interest of innovation as a "good thing", short-term disruptions with the accompanying legal violations should also be possible.

Moreover, Silicon Valley could also evidently rely on the legal analysis already described. At least in Germany, the hypothesis of "training generative AI models = TDM" was put forward shortly after the capabilities and functionality of generative AI became known. More and more publications gradually followed suit.

"So much for the dream of the public domain"

On the one hand, we are dealing with innovative start-ups in the AI industry, which the EU's AI Act obviously wants to defend –, especially if they come from Europe. On the other hand, it looks very much as if only market giants such as Meta, Google or Microsoft really benefit seriously from free training because they are likely to dominate the market. Why is there such a disparity when Brussels is trying to contain the giants with DMA and DSA at the same time?

Dornis: This is indeed an almost paradoxical situation. The EU is fundamentally committed to taking the lead on digital issues. By protecting individual rights and minimum standards, the aim is to create a space and, above all, markets in which competition functions fairly and with respect for all interests involved. With consistent regulation in this sense, it is hoped that the next step will even be to influence other legal systems, in particular the USA.

Put simply, if global players adhere to EU standards because the EU is an attractive market, this will also increase the level of protection in other countries and perhaps even influence legislation. While this is euphorically glorified as the "Brussels effect" in almost all areas –, including the digital marketplaces –, authors do not appear to be among the groups to be protected. At the moment, however, they are actually the ones whose rights are "falling by the wayside" during AI training.

The ChatGPT inventor OpenAI will say that they already have to raise enormous sums of money on the road to general artificial intelligence – Sam Altman is currently collecting billions. Is there any money left over for the creatives?

Stober: I think that's a strange idea. The level of remuneration for creatives whose data was used to train this commercial generative AI should not depend on how much money the company has raised. It seems more sensible to me to base it on turnover. But I'm really not an expert on this. My focus here was on clarifying the technical details.

Purely AI-generated content has a completely different problem: it cannot actually be protected by copyright. Are we now entering the golden age of the public domain, created by media corporations?

Stober: I wouldn't expect profit-oriented media groups to make the output of their generative AI models available free of charge. I can't say anything about the legal issues here.

Dornis: Even lawyers can say little about the value of AI-generated content, and certainly nothing about the question of whether we consider self-replicating AI creativity to be desirable. However, human creativity must always be taken as the starting point. Even the most highly developed AI cannot create anything out of nothing. It will therefore never work completely without incentives for human creatives. So much for the dream of the public domain.

Could creatives become mere henchmen of AI in the future so that generative content can be protected after all? There is currently the hype profession of prompt engineer, but generative AI itself could quickly replace it.

Stober: I don't actually consider prompt engineering to be a job with good prospects for the future. However, I find the image of "AI's sidekick" a little unpleasant.

AI is nothing more than a tool, albeit a very powerful one. It is people who decide how to use it. In this respect, the creatives here would then be the henchmen of media corporations in the assembly line production of content.

What could a form of fair pricing for AI training look like? Buyout contracts are probably rather bad for creatives.

Dornis: As a lawyer, I'm not sure what the "fair" price is. However, the law must of course provide functioning mechanisms and processes for compensation. The law currently in force cannot do this effectively. Moreover, the discussion at the moment is rather foggy, especially when it is argued that only low royalties can be achieved if creative works are licensed for AI training. This is simply wrong: where there is no market because everything is available for free (as is the case at the moment), no market prices can develop –. In any case, these are far below what a functioning license market for training data would generate. In any case, the answer to questions about future remuneration should not begin with a reference to the status quo.

Stober: In addition, a social debate on the general value of data would certainly not be a bad idea. The fact that it is currently being collected on a grand scale by AI companies with a certain self-service mentality and monetized in products and services based on it is perhaps not in the interests of the general public. And this doesn't just affect creatives, but everyone on the Internet.

How do you think the situation will develop now?

Dornis: There is already extensive litigation in the USA and UK. This will also happen in Germany and on the European continent. We will then see how quickly the national courts refer the relevant questions to the European Court of Justice. However, this will take some time. It will therefore remain exciting.

Stober: The topic will certainly keep us busy for quite some time. I hope that we can contribute to an objective debate with our study. (anw)