What’s the status of AI and copyright?

Do androids dream of electric llamas?

It’s been an interesting couple of years here in Llama Towers. While AI has long been a subject of interest to this blog, it’s undeniable that the subject has exploded recently due to the growing number of generative AI tools made available to the wider public. Only two years ago many of us were playing with Craiyon, and rumours of an incredible tool called DALL·E 2 were starting to make waves. ChatGPT wasn’t even a thing, and Midjourney and Stable Diffusion were still a couple of months away.

Think of everything that has happened since. The flood of tools and apps, the lawsuits, the inclusion of generative AI tools in large parts of the tech environment. Who could have thought it possible? Well, plenty of people did. While there’s quite a lot of AI hype going around, it’s undeniable that we are living in a different world, and models are just improving all the time. So this is as good a time as any to take a look at the state of play, and take stock of what is happening in AI and copyright, and perhaps to look forward to the next couple of years.

The lawsuits

Perhaps the most talked-about aspect of AI copyright has been the lawsuits from various copyright holders against AI developers and tech giants. It’s not the objective of this blog post to enumerate all of the ongoing cases; there are places where you can follow the current state of the ongoing litigation. However, a few trends are starting to emerge that should give us an indication of things to come.

The first is that the lawsuits are proceeding slowly, as expected, but personally I’ve been surprised that at least at the time of writing there haven’t been more out-of-court settlements. It is perhaps too early for that, but I would expect that a good number of these cases won’t make it to trial, either due to settlement or because some of the companies involved may go bankrupt. In the coming year we should see more resolutions in this respect, probably a few will be negotiated or dropped.

As for the outcome of the lawsuits, this is evidently more difficult to predict. There are a variety of claims in the ongoing litigation that have been dismissed, but for the most part the cases will be decided on the question of whether training an AI with unauthorised copies of works is fair use (in the US cases). I’ve always thought that this could go either way. It is perfectly possible that we may get different decisions in some of the cases, with some courts declaring fair use, while others declaring that this is infringement.

What I fully expect is that regardless of the outcomes, the cases that get a decision will almost certainly be appealed, so we’re in for years of further litigation without any answers. We should also add to the uncertainty the fact that we have yet to see cases brought by the music industry and the film industry against the fledgling audio and video generative AI startups. This suggests we should expect years and years of lawsuits to come.

Effect of the lawsuits on the AI industry

While the exact outcome of the ongoing litigation is uncertain, I’m more confident in making predictions about their effect on the industry as a whole: I don’t think that the cases will have a damaging effect on AI development. It’s certain that some individual companies may suffer substantial monetary awards, but the industry as a whole will survive regardless. In fact, I’m prepared to go one step further and declare that even if every single company in existence were to go out of business tomorrow, AI development would continue unabated.

The reality is that despite the hype (and perhaps because of it), AI is big business, and its inclusion in a large number of applications appears to be inevitable. To me, the true indication that AI is here to stay is its inclusion in future Apple products. It would be foolish to underestimate the power of the hardware giant in moving trends forward, and once generative AI is seamlessly included in everyone’s phones, I simply cannot see how copyright lawsuits would stop this momentum.

What is likely to happen is a combination of factors that will eventually stabilise the tech marketplace. I envisage a future in which most cases have been decided one way or another, and most generative AI takes place either under licensing agreements, or in compliance with existing (or future) exceptions and limitations. Markets abhor uncertainty, so it is only natural that some resolution will take place that will allow consumers, rights-holders, and developers to co-exist in relative peace.

Moreover, the sheer pace of AI innovation suggests that by the time these legal battles are resolved, the technology landscape may have shifted dramatically. We’re likely to see new AI models and techniques emerge that could potentially sidestep some of the current legal concerns. For instance, we might witness the rise of AI systems trained exclusively on public domain works or specially curated datasets, circumventing many of the copyright issues we’re grappling with today.

Another factor to consider is the global nature of AI development. While much of the current legal focus is on the United States, AI research and implementation are happening worldwide. Countries with different legal frameworks and attitudes towards copyright may become new hubs for AI innovation, potentially shifting the balance of power in the tech world. This global dimension adds another layer of complexity to the issue, but also ensures that AI progress is unlikely to be halted by legal challenges in any single jurisdiction. All US cases could be decided against AI companies, and that would still not stop the international development of AI, some countries could even take advantage of that fact and become training safe havens.

The state of regulation

The main developments with regards to regulation have already taken place in the shape of the European AI Act (more detailed coverage of the Act is available here). While the effects of this regulation on copyright issues are either too far-reaching or not enough, depending on your outlook, what is certain is that tech companies will spend the next couple of years preparing for its implementation.

The main provision in the Act is the transparency requirement. At the risk of oversimplifying, the Act will require providers of generative AI tools to prepare transparency reports that will list the works used in training. While we do not yet know just how detailed these reports will have to be, it is almost certain that they will at least have some effect on AI developers. Similarly, one of the most controversial aspects of the AI Act will be its expected substantial effects outside of European jurisdictions. We’ll also have to see the details of this, but at present, the Act will apply to companies which provide AI tools to European citizens.

Regulatory action across the pond appears to be easier to predict. While the FTC has been making waves with regards to generative AI, the action undertaken up to now has been limited. The chaotic state of US politics makes it difficult to discern what will be the next activity there, so for now we can mostly expect the lawsuits to provide most of the regulation in the US.

It’s worth noting that the regulatory landscape is likely to evolve as rapidly as the technology itself. As we’ve discussed earlier, the global nature of AI development means that we may see a patchwork of regulations emerge across different jurisdictions. This could lead to a situation where AI companies have to comply with often conflicting international rules, potentially favouring larger corporations with the resources to ensure compliance across multiple regions. Smaller startups and innovators might find themselves at a disadvantage, potentially stifling some avenues of AI development while inadvertently promoting others.

Moreover, the interplay between regulation and ongoing litigation will be crucial to watch. As courts grapple with copyright issues in AI training, their decisions could influence future regulatory approaches. Conversely, new regulations like the EU AI Act might shape the interpretation of existing laws in court cases.

The authorship question

Perhaps one of the most surprising developments in AI copyright has been the relative lack of attention to the authorship question, that is, the issue of whether works generated by an AI are subject to copyright protection. This has been an important issue for copyright nerds such as Yours Truly, but has failed to generate much traction in the press, as well as not being a particularly litigated area of copyright, with a few exceptions.

The reason for this may be a combination of factors. The first is that AI-generated works haven’t reached the professional level yet, and while some artists have adopted AI tools, it still remains a relatively small niche with limited commercial value. This is also an area in which the over-abundance of AI-generated works translates into little commercial value for AI outputs. Why pay for some generated artwork, audio, or text, when you can just produce your own? This is what a post-scarcity economy looks like, and one of the results is that those using the tools aren’t interested in copyright protection of their own works. Personally, I’ve generated thousands of images, and I’m not remotely interested in protecting any of them.

The result is that the copyright status of most AI-generated content has gone unchallenged in courts, and a parallel economy has emerged in which AI enthusiasts produce works with little worry about whether someone will copy their creations.

This may change in the future when AI tools get better, and more AI-generated works make their way into mainstream media production and large copyright projects in the shape of images, scripts, music, and animation. Until that happens, the authorship question will continue to be eclipsed by the copyright infringement issue.

Concluding

Artificial intelligence is here to stay, fanciful ideas of it being destroyed by copyright lawsuits aside.To some people this statement will be triggering, an admission of defeat, a herald of a dystopia where the machines have won and the human spirit has been extinguished. I do not know what the future will look like, but at least I’m confident that copyright won’t be the one shaping it, and we will have to find our answers around important questions in other ways. I tend to be an optimist and a believer in the human spirit, so I would at least think that we’ll manage to find a way through.

One thing is sure though, the only ones to benefit in the short term are likely to be copyright lawyers.