AI and Copyright in the UK – TechnoLlama

(This is the text of a recent short piece I published on SSRN, it’s short enough to be turned into a blog post, but it’s written in a more formal style than what habitual users are accustomed to).

 

Copyright has been at the forefront of many of the policy conversations surrounding artificial intelligence (AI) in recent years, particularly since the rise of generative AI. It is therefore no coincidence that the United Kingdom stands at a pivotal juncture in this very topic. The unrelenting advancement of AI presents a great opportunity to secure a position at the forefront of this new technology. To seize this moment, the nation must proactively shape a regulatory environment that fosters innovation while respecting the important creative industries. Central to this is the urgent re-evaluation of the UK’s copyright framework, a process already underway with the government’s recent consultation on AI and copyright. This article argues that a progressive reform of the text and data mining (TDM) exception is an important enabler for making the UK a globally competitive destination for AI development.

The government’s consultation has rightly identified several key areas for consideration: copyright authorship for works generated by AI, the potential for licensing regimes, enacting transparency requirements in AI training, and a vital revision of the exceptions and limitations surrounding text and data mining. It is on this last point that the UK’s future as an AI powerhouse may hinge. By embracing a forward-looking TDM exception, with at least a robust opt-out mechanism for creators, the UK can signal that it is open to investment in this vital sector, as well as being able to stimulate homegrown innovation in AI technologies.

The need for change

It is often difficult to remember that the current rise of generative AI has only taken place in the last few years, arguably with the launch in 2022 of ChatGPT. Official estimates in 2023 suggested that the UK’s AI sector employed 50,040 people but the UK AI Opportunities Action Plan notes that AI capabilities are “developing at an extraordinary pace”. AI has also become an important part of the government’s economic strategy looking forward to 2035, betting on further increase in this industry. The UK already has a strong foundation in AI, with a vibrant ecosystem of startups and academic research institutions. The AI sector already contributes billions to the UK economy and is a powerful engine for job creation.

However, the existing success stories do not guarantee future leadership in the global AI research race. Countries like the US and China have been moving forward as hubs of AI innovation, and in order to keep up the UK will have to tackle several issues. It is no secret that AI models require access to vast amount of data for training purposes. The legal framework governing the use of this data is critical in determining the nation’s competitiveness in the AI race. As other nations have more permissive regulations on the books, this places the UK at a relative disadvantage unless this is remedied.

A recent report commissioned by the Computer & Communications Industry Association (CCIA) helps to illustrate the economic case for reforming the copyright exception regime. The study analyses the potential impact of different copyright scenarios on AI and investment in the UK. On the one hand, a commercially viable TDM exception could see the UK’s share of global AI-related investment reach £3.2 billion GBP annually. Conversely, maintaining the status-quo or adopting a more restrictive approach could result in a potential loss of between £0.8 billion and £1.8 billion in annual investment.

Aligning with global trends

Given the options at hand, the most likely exception scenario is the government’s preferred option in the ongoing consultation, namely one encapsulated in Option 3: “A data mining exception which allows right holders to reserve their rights, underpinned by supporting measures on transparency”. This option allows for a clear opt-out for rightsholders, and it is already in place in other jurisdictions. This would also bring the UK into closer alignment with the direction of travel in other major economies, notably the European Union and, in a different but functionally similar way, the United States.

The EU’s Digital Single Market Directive (DSM), specifically Article 4, provides for a TDM exception for commercial purposes as long as they respect any express reservation of rights by the copyright holder, and this has to be done through machine-readable means. This “opt-out” model seeks to strike a balance between enabling large-scale data analysis for training, and allowing creators to maintain control over the use of their works. The implementation of this opt-out mechanism has presented some practical challenges, and the courts have been engaged in attempts to explain the limits of the exception. However, at least for the foreseeable future the opt-out regime will stay in place in the EU, especially after the publication of the final draft of the AI Act’s General-Purpose AI Code of Practice, which cements the existence of the reservation of rights regime.

Crucially, there is no evidence to suggest that the opt-out approach has had a detrimental effect on the creative industries in the EU. While there are justified issues to address with the opt-out system, particularly regarding the nature of the reservation of rights, the creative industry in Europe has continued to flourish despite the shock of the COVID pandemic. The EU’s creative economy is a heavyweight of €600 billion EUR of annual value-added and roughly half-a-trillion in yearly turnover. After the 2020 downturn, the industry is growing again: 2022 turnover was about € 60 bn higher than in 2020, and more recent value-added estimates suggest it continued to expand through 2023-24. This appears to be in line with other developed economies. With this in mind, the opt-out framework has provided a degree of legal clarity that was previously absent, allowing AI developers to proceed with a greater understanding of the applicable rules.

Japan has also adopted a notably AI-friendly copyright law, which permits the use of copyrighted works for the purpose of information analysis without the need for a licence, provided it does not “unreasonably prejudice the interests of the copyright owner.” The Japanese exception for TDM has positioned them as an attractive location for AI research and development, and again, there is no substantive evidence that this has led to a decline in its vibrant creative sector. While the reach of the existing exception is debatable, it is clear that its existence helps to debunk any argument that the adoption of TDM exceptions affects negatively the creative sector.

In the US, a different but equally important trend is emerging through the courts. Fair use is an open-ended exception that is broader and more case-specific than TDM exceptions discussed above, and this has been a key feature in recent high-profile copyright infringement cases brought against AI developers like OpenAI and Anthropic. While these cases are ongoing, there have been two early decisions pointing towards the acceptance that in some circumstances AI training can be classified as transformative, and therefore falls under fair use. There are over 40 cases being litigated at the moment, so it is likely that these will result in a variety of decisions, probably some will find for fair use, while others may find infringement occurred. But one thing is clear, the question will remain open for years to come, and there is always a strong possibility that training is fair use. The prevailing winds in US jurisprudence appear to be blowing in a direction that is favourable to the type of data processing inherent in AI model training.

This means that if the UK adopts a TDM exception with an opt-out clause, it would be seen as recognition of the growing international consensus that recognises the need for a more flexible approach to copyright in the age of AI.

Territoriality and jurisdiction

Copyright law is strictly territorial by its very nature, it is common to see experts comment that there is no such thing as international copyright law, just an system of national legislation and a system of treaties that allow for common recognition of rights. A copyright from a UK citizen is enforceable in the UK, and while it is possible for a rightsholder to exercise their rights in other jurisdictions, the territorial nature of copyright remains. This is extremely important issue with regards to AI because of the international nature of the technology. Data is gathered, the origins of which can be a number of sources, some may be protected by copyright and some may not; this data is then used for training using computing resources located in one country. These are used to train models that are the basis for AI tools and platforms that can be located in one country and used in another.

The practical effect of the territoriality principle is that a work from a UK rightsholder may be used for training in a country that allows it to take place without infringing copyright law, such as the situations described above, this could generate a conflict between the owner of a work and an AI developer, both could be operating under different expectations, the author could expect any training to be infringement, while the developer would operate thinking that their actions fall under fair use or fair dealing.

A lot of the discussion surrounding the UK government’s consultation on AI and copyright has been centred on the question of how it would affect the nation’s creative industries, but something seems to be missing from this debate. It is common to hear that any sort of exception would be detrimental to UK rightsholders, and it would erode their rights. The reality is that works from UK authors are already being used in AI training in jurisdictions that have permissive copyright regimes, or where the question has not been decided by the courts. If that is the case, then UK rightsholders could attempt to sue for copyright infringement in this jurisdiction. But that presents its own issues.

A recent example of this has been the case of Getty Images v Stability AI, which is ongoing at the time of writing. Stability is a UK company which also operates in the US, and their models were trained in the US. Getty Images is a stock photography company that has legal personality established in the UK, so they sued for copyright infringement in England. While the case has not been decided yet, an important part of the litigation was centred on the fact that no training took place in the UK, which may have prompted Getty to drop their primary copyright claims. The argument boils down to this: if there was no training in the UK, there was no infringement.

A decision to tighten copyright law in the UK in a way that makes AI training prohibitively difficult or legally precarious will not stop AI models from being trained on UK-created content. It will simply mean that this training happens elsewhere. UK-based AI developers would be placed at a significant competitive disadvantage, unable to compete with their international counterparts who operate under more favourable legal conditions. The result would be a chilling effect on domestic AI innovation, with investment and talent flowing out of the UK to more welcoming jurisdictions.

There is also the very real risk of “forum shopping” in copyright litigation. This is the practice of strategically choosing the jurisdiction in which to bring a lawsuit in order to take advantage of favourable laws. If the UK were to become an outlier with an overly restrictive copyright regime, AI developers might be hesitant to offer their tools and services in the UK for fear of being dragged into costly and protracted legal battles. This could lead to a situation where UK businesses and consumers are denied access to the latest AI tools that are readily available in other parts of the world. The UK would become a digital island, cut off from the mainstream of technological progress.

The way forward

The current debate over AI and copyright is not merely a legal squabble; it is an important discussion about the future of the UK’s economy and its place in the world. The choice is between a future where the UK is a dynamic hub of AI innovation which attracts investment and talent from around the globe, and a future where it lags behind, held back by an outdated legal framework.

An updated TDM exception which incorporates a workable opt-out for creators is the cornerstone of a pro-innovation approach. The exception presented in the consultation acknowledges the potential of AI, while at the same time it respects the rights of creators. It aligns the UK with international best practice and provides the legal certainty needed to unlock the economic potential of the AI industry.

The government’s consultation is a step in the right direction. By embracing a forward-thinking and balanced reform of its copyright law, the UK can send a clear signal to the world that it is open for business, ready to innovate, and determined to be a leader in the artificial intelligence revolution.