AI and Copyright – A Consultation by the UK Govt

What happened

• A wide reaching consultation has been launched by the government on AI and copyright which aims to finally settle the ongoing debate surrounding these
• The consultation asks for feedback on a range of critical areas including text and data mining exceptions (TDM), the territorial scope of UK copyright law, collective rights management, transparency requirements and the issue of protecting AI assisted/generated works
• AI has an increasing influence on the creative industries and the UK must catch up with the EU on many issues and balance the rights of creators and rightsholders with the need to allow AI to enable innovation
• AI models are often trained on vast datasets that include copyrighted materials, such as books, articles, music and art. This training process can take place without explicit permission from the original creators, leading to potential copyright infringement issues
• The concept of ‘fair use’ allows limited use of copyrighted material without permission under certain conditions, but applying fair use to AI is complicated
• If AI generated content closely resembles existing copyrighted work, it might be considered as a derivative work, which also needs permission from the original copyright holder
• If AI is used to generate content, this can affect the market for the original work. Where AI-generated content competes with or devalues the original work, it could harm the market value of that work

Why it matters

• Getting the balance right between using AI and adhering to copyright laws is important, because any copyrighted work will have a human creator
• The question is, who owns the rights to the AI generated work?; is it the developer of the AI, the user or the AI itself. Current laws do not however recognise AI as an author
• Policymakers seeking a balance, propose to adopt a TDM exception to copyright allowing AI developers to use copyrighted works for training purposes. This would be coupled with an ‘opt out’ rights reservation package for rightsholders and greater transparency requirements
• Transparency is crucial in the government’s proposal. It will be necessary to (i) disclose specific works/datasets used in AI training (ii) disclose web crawler details including ownership and purpose (iii) keep records  to demonstrate compliance

Next steps

The government also plans to clarify copyright law further and this marks one of the first significant differences between EU and UK copyright law.  It includes clarifying the scope of the ‘temporary copies’ exception for AI developers and its international application. 

The consultation closes on 25 February 2025.

Details can be found at 

Copyright and Artificial Intelligence – GOV.UK

What happened

• The UK government announced ambitious plans in January 2025 for a number of hyperscale data centres of 100MW + to be developed
• It is hoped that the UK will become a leader in the global data centre industry. The plan is for a number of AI Growth Zones (AIGZs) to be created
• Planning permission grants are the key to this and these will need to be sped up
• Currently there is a more favourable regime for a ‘national infrastructure project’ – however, a data centre does not count as a ‘national infrastructure project’ as it does not relate to transport, waste, water or energy under the criteria of the Planning Act 2008.
• If new legislation is introduced, then hyperscale data centres might be included as ‘national infrastructure projects’
• Additionally, a permitted development right to convert a brownfield site into a data centre might be introduced, and this might speed up the planning process as well as to remove a local planning authority’s ability to veto a project
• Important to note is that data centres have a high energy requirement – running a ChatGPT query needs 3 watt hours of electricity – which is 10 times that needed for an ordinary Google search
• 12 watt hours are needed to generate a single image using AI – about the same time to fully charge a smartphone
• Data centres need better access to power, but electricity distributors are already very overstretched so it might be difficult to obtain enough power supply
• An alternative might be for data centres to procure their own power. On-site wind and solar generation might be useful, in certain areas but for hyperscale sites this is not sufficient and nuclear power might be the way forward

Why it matters

• If the UK is to become a leader in the world of AI, it is essential that hyperscale data centres can be developed
• SMRs (small modular reactors) are a good option for data centres because (i) they can generate a huge amount of stable, reliable, always on, energy from a small amount of land (ii) it is difficult for data centre operators to predict how much power they will need in 5 or 10 years, and SMR models can be scaled up as the data centre needs
• There are currently no small modular reactors (SMRs) on land that are used commercially. The technology is being developed simultaneously in several countries (the first one is due in China)
• If SMRs are to be used for data centres, the government will need to ensure they can keep pace with the huge rise in demand for AI services

Takeaways

A clear link between the future of AI and nuclear energy has been set out.  It is likely that demand for data centres is the catalyst behind the SMR development that the world needs.  

We will need to watch this space closely…….

What happened

• There are many benefits in using the cloud such as greater flexibility and cost efficiencies
• However, cloud-based arrangements are vastly different, in both nature and risk profile, from more traditional forms of technology procurement
• Under a traditional software licensing model, data processed by the software application remains under the control of the customer
• Where software is provided on a cloud-based software-as-a-service (SaaS) basis, data processed by the application will typically be hosted and stored by the software provider, so it is no longer under the control of the customer
• This creates a number of new risks relating to information security, operational resilience and data protection
• If the data being hosted by the software provider is the only, or main, instance of that data – then the customer will need to consider the potential impact of data loss and how to mitigate that risk
• General and sector-specific rules around operational resilience and information security may also impact SaaS solutions. For example, the new EU regulation around financial services operational resilience (known as DORA), must be considered where a service recipient or service provider falls within its scope

Why it matters

• When using cloud services, businesses must fully understand the risks involved and how they impact their organisation
• If the data being processed, includes personal data then both the customer and supplier will need to ensure relevant data protection laws are being complied with
• An important first step is to complete a Data Protection Impact Assessment to determine the risks
• Laws around international data transfers must also be considered, where the data is hosted and stored in another jurisdiction

What next?

The cloud is here to stay and has many benefits as a technology delivery model, however there  does need to be an understanding in how both customers and software providers view the procurement of cloud-based solutions to better align with how this delivery model works.

Software providers must acknowledge their different roles and responsibilities when moving from a traditional software licensing model to a cloud-based SaaS delivery model and tailor their contracts accordingly

The government produced information on ‘multi-region cloud and software as a service’ on 5 February 2025 which can be found at Multi-region cloud and software-as-a-service (HTML) – GOV.UK