Specifically, the UKIPO asked whether the existing text and data mining exception in UK copyright law is adequate to give AI developers access to the data they need. An answer to that question is anticipated from the government within weeks.
The current position
Text and data mining (TDM) is the use of automated computational techniques to analyse and copy large amounts of information to identify patterns, trends, and other useful information. TDM may be used to develop and train AI.
Section 29A of the Copyright Designs and Patents Act 1988 sets out a text and data mining exception to copyright. This provides that TDM will not infringe the copyright in a copyright protected work provided an AI developer has lawful access to the work – for example, via a subscription or permission in terms and conditions – and the TDM is for the purposes of non-commercial research. Contract terms that stop developers relying on the TDM exception are unenforceable and the law also requires acknowledgement of both the works mined and the rights holders unless this is impractical.
The TDM exception has significant limitations, not least the requirement that the TDM must be for non-commercial purposes. This is problematic in the context of AI, as most transformational AI is developed with a view to future commercialisation. In addition, the exception only applies to works protected by copyright and does not extend to database rights.
This means that AI developers undertaking TDM for commercial purposes must limit TDM to works not protected by copyright or database right. This is likely to severely reduce the data available, and mining of the text may in any event be prevented by other contractual or legal restrictions, such as data protection requirements. Otherwise, AI developers will need to seek a licence from the rights holders.
Sometimes a licence may be freely and easily available, for example where the rights holders have chosen to use generic open licence conditions such as Creative Commons or Open Government Licence or made their data available on a subscription basis with the uses to which the data can be put set out clearly in the related terms and conditions. However, in some situations individually negotiated licences may be needed. This can create difficulties where the rights holders cannot be identified or successfully contacted, and a licence fee is likely to be payable.
Options for change
In its consultation, the UKIPO asked for views on whether the TDM exception should be extended. It set out five potential options. The first is to maintain the status quo, although clarification would be provided on what constitutes a non-commercial purpose. The UKIPO said that choosing this option would be justified if the current exception proves not to be an impediment to accessing material for TDM, in particular when training AI systems.
The second option would also involve no change to the existing TDM exception but would encourage in parallel an environment of best practice licensing for TDM through education and the introduction of model licences or codes of practice.
In its discussion of this option, the UKIPO highlighted the benefits of licensing in the TDM space, including the greater legal certainty around the mining permitted, the ability of rights holders to benefit from licensing income and the possible encouragement to some rights holders to provide value-added data-mining products. However, the UKIPO also acknowledged the corresponding licensing costs to AI developers as well as the transaction costs involved in locating rights holders and negotiating the licence agreement.