Training AI models: UK Government consults on proposals for a copyright data mining exception with rightsholder opt-out

AI Model Training

Currently, the UK’s copyright exceptions in the area of text and data mining and ownership of computer generated output differs from the EU framework, in part, because the UK did not (following Brexit) implement the EU DSM Directive, which contains a text and data mining exception for any purpose, subject to rightsholder reservation of rights. The government’s proposals would therefore broadly align the UK framework more closely with the EU framework under the DSM Directive, with the addition of transparency obligations on developers.

Whilst the data mining exception with rightsholder opt-out is the government’s current preferred option, the government is consulting on various options, from doing nothing, strengthening copyright so that developers would have to obtain a licence in all cases, and a broad data mining exception with no rightsholder opt out (which the previous government sought views on and consequently withdrew proposals on). In relation to the data mining exception proposals, the government is seeking views on various aspects of this, including: what level of transparency would be required by developers; the standardisation of technical tools for rightsholders to use to reserve rights; how the government should support licensing, in particular whether collective licensing would be useful and whether labelling of AI outputs would be beneficial. 

Ownership of AI output and emerging issues

Whilst the focus of the consultation is AI model training, the government also seeks views on ownership of AI output, in particular whether the UK’s specific rules on ownership of computer-generated works should be removed, as unworkable, and contradictory to EU copyright laws. Other emerging issues the government is seeking input on include the challenges of AI-generated voices, images and likeness and the use of synthetic data to train models.  

Next Steps

In total, there are 47 questions to be answered, many of which relate to the government’s proposals for a data mining exception with rights reservation, including how it would work in practice. We have been following the developments in this area closely and will continue to report on them. If you would like help with responding to the consultation please do contact our London Copyright Practice. 

Authored by Penelope Thornton.