UK Government consults on copyright and Artificial Intelligence

January 15, 2025
Mike Knapper Michael Sinclair Rosie Nance
Mike Knapper Michael Sinclair Rosie Nance
Posted in Artificial intelligence Intellectual property

The UK government has issued a new consultation, Copyright and Artificial Intelligence, under which it is proposing to address a tension that exists between:

  • the interests of rightsholders (such as creative and media organisations, or indeed anyone publishing content on the Internet), who are concerned to ensure that their copyright works cannot be used by AI businesses to train AI without their permission (and potentially payment); and
  • the commercial imperative of AI businesses to be able to use such works to train their AI systems in the UK without infringing copyright (such businesses arguing that the inability lawfully to use such training data could impede the growth of the AI industry in the UK).

Business users of AI are also concerned about a perceived lack of transparency from AI developers about what content has been used to train their AI, giving rise to potential user infringement exposure (see our publication, Generative AI: a global guide to key IP considerations, for a discussion of these risks).

To address these considerations, the UK government is proposing:

  • A new exception to copyright law for text and data mining (TDM): 
  • AI developers would be able to train AI on material available on the Internet (or accessed, say, via subscription) to which they have lawful access (that is, not gated content, nor subject to a contractual restriction on use, etc) for any purpose (including a commercial purpose - something not currently provided for under existing TDM provisions under the UK’s Copyright, Designs and Patents Act 1988 (CDPA 1988)), but only to the extent that right holder has not reserved its rights.
  • Right holders would be able to control the use of their works using technology (perhaps a machine-readable format, such as the robots.txt standard, metadata, or a new standard) and seek payment through licensing agreements (including collective licensing). 
  • This approach is similar to the EU’s exception for TDM in Article 4 of the Digital Single Market Copyright Directive (Directive (EU) 2019/790).
  • Increased transparency: Article 53(1)(d) of the EU’s AI Act recently introduced a requirement for general purpose AI systems that AI training sources are reported (providers of general-purpose AI models are required to make publicly available a “sufficiently detailed summary” of training content). The UK government makes clear that it will not introduce a new TDM exception without robust transparency measures and seeks views on these.  The consultation notes that these might include requirements to disclose the use of specific works or datasets, details of web-crawlers used (and the purpose for which content is being crawled), to keep records, to provide certain information on request, or to evidence compliance with rights reservations. The consultation does not favour any particular approach.
  • Contracts and licensing: The UK government also seeks views on whether creators and performers have sufficient control of the treatment of their works when industry expectations are that they should cede much control (including in relation to AI use) in industry-standard contracts.  The UK government considers that an enhanced TDM exception of the type proposed may need to be supported by government-facilitated collective licensing arrangements.
  • Temporary copies exception: The CDPA 1988 includes an exception that permits temporary copies to be made during technological processes, such as a copy in a browser cache. It is not clear whether this exception applies to the training of generative AI models.  The UK government seeks views on whether this should be clarified (presumably by making it clear that the exception does not apply to training AI).
  • Research and innovation:  The CDPA 1988 provides for a specific exception for TDM for non-commercial research.  The equivalent EU provision permits commercial research and extends to databases as well as copyright works.  The UK government is seeking views on whether the UK exception should be similarly extended.
  • Computer-generated works: The CDPA 1988 provides copyright protection for computer-generated works.  There are various technical arguments as to the extent to which this protection applies to AI outputs.  The UK government consultation asks whether the provision should be clarified or removed, noting that most other “leading AI nations” do not have an equivalent position in their laws (some countries such as India, Ireland and Singapore have an equivalent), but states that if the consultation does not reveal sufficient evidence of the positive effects of this provision, the government’s preference is to remove it.
  • Infringing outputs: AI trained on copyright works could produce infringing outputs (see our publication, Generative AI: a global guide to key IP considerations, for a discussion of this risk).  The UK government says that it expects that greater control and transparency over the inputs to AI models (such as the TDM measures it proposes) will reduce the risk of AI producing infringing outputs.  Nevertheless, it seeks views on whether copyright law is deficient in relation to AI producing infringing outputs.
  • AI output labelling: The EU AI Act lays down transparency rules for content produced by generative AI, requiring providers to ensure AI outputs are marked in a machine-readable format and detectable as AI generated or manipulated.  The UK government seeks views on this approach and whether generative AI outputs should be labelled as AI generated.
  • Digital replicas: Digital replicas (or deepfakes) replicate a person’s voice, image, or personal likeness. While not consulting on specific proposals for so-called “personality rights”, the UK government is consulting on whether individuals have sufficient control over use of their likeness under existing UK laws, such as the law of passing off.

Next steps

The consultation closes on 25 February 2025.  The UK government says it will use the information it receives to help it design policy.  It is not clear whether this would immediately result in draft legislation, although the consultation clearly envisages that regulation will be needed in many areas.

Our take

In consulting on AI and IP rights, the UK government is seeking to strike a balance between competing commercial interests. On the one hand, the UK creative sector’s contribution to the UK economy is significant (£124bn per annum according to the consultation), but the government clearly sees the current limits on AI companies’ ability to train their systems in the UK as putting the UK AI sector at a disadvantage to their international competitors.  Explicit reference in the consultation to EU, US and Singaporean AI and TDM legislative initiatives (among others) underscores the UK government’s awareness of the fact that it is competing in a global market for AI investment where AI industry stakeholders inevitably factor in the regulatory environment in their investment decisions.      

While there will be strong feelings on both sides, and we note that the 2022 proposal to introduce a commercial TDM exception was abandoned by the then-government due to the concerns raised by the creative sector, whatever the outcome of the consultation, the fact that the government is proposing that any changes are made through legislation, rather than through a voluntary code of practice (which was the objective of a prior 2023 IPO-led working group) is to be welcomed, because of the clarity that this would bring.