AI and Copyright - the Kidron Amendments

The UK's House of Lords (the upper chamber in the UK's Houses of Parliament) has voted to add what are known as Baroness Kidron's amendments to the Data (Use and Access) Bill. This amendment would, if it becomes law, require the Secretary of State to force AI developers to publish information about the works used to train their models.

What is the Bill?

The Bill was introduced in October 2024 with the purpose of "unlock[ing] the secure and effective use of data for the public interest, without adding pressures to the country’s finances". It has a number of objectives, many of which are not particularly relevant to IP lawyers, e.g. speeding up access to data for the health service, making it easier for law enforcement to access data, and enabling the use of digital identities instead of physical identity documents. However, the Bill has been used as a mechanism to address some perceived ills of the growth of AI services, including the use of copyright works for training the underlying models and for dealing with the creation of deepfake 'intimate images' of adults (the creation of intimate images of minors already being illegal). 

What is the amendment?

Baroness Kidron has been the figurehead in the House of Lords fighting for the UK creative industries for a while, and has been using this Bill to do so. An earlier amendment she proposed would have required the Secretary of State to "make provisions clarifying the steps the operators of web crawlers and general-purpose artificial intelligence (AI) models must take to comply with United Kingdom copyright law, including the Copyright, Designs and Patents Act 1988."

That amendment did not make it into the Bill that was approved by the House of Commons earlier this month. The Commons instead proposed a requirement that the Secretary of State "prepare and publish a report on the use of copyright works in the development of AI systems" within 12 months, at the same time as publishing a report on the economic impact of the four policy options put forward in the Copyright and AI Consultation Paper.

The Lords has voted, by a sizeable majority, to add an amendment requiring the Secretary of State to make provisions that:

• AI developers must provide copyright owners with information regarding "the text and data used in the pre-training, training, fine-tuning and  retrieval-augmented generation in the AI model, or any other data input to the AI model."
• AI developers must give details of the bots used to collect the information used by the services, including the name of the bot, the legal entity responsible for it, and the purposes for which it was used. 

What is the significance of this?

The first thing to say is that the Lords proposed this Bill and the amendments, but the Lords does not have the power to force it through. The Commons will need to approve the Bill for it to become law, and it is fair to say that the Commons is divided on the promise of AI. It is notable that the Lords amendment was passed primarily with non-Labour Party support, with Labour peers largely voting against it. The Labour Government is keen on promoting growth, but is wavering on whether growth is best supported by promoting the interests of the creative industries or of promoting the interests of the tech sector. The Labour Party is of course in government, with a large majority in the Commons, and it seems likely that the differing makeup of the Commons will prevent the amendment making it through. 

The second is that the amendment does not, to this Kat, address the fundamental reason for the lack of enforcement of copyright by UK rightsholders against AI developers. Most development appears to be taking place outside the UK - a key argument being put forward by Stability AI in defence of a claim by Getty Images. If the reproduction is not taking place in the UK, and whatever outputs are not infringements of the original training data, it may be that there is no infringement taking place in the UK at all. While litigation may be possible in the USA or other jurisdictions, the law is uncertain, costs can be prohibitive, and the value of any individual claim uncertain (and likely small). 

The Getty Images v Stability AI case is due to go to trial on 9 June 2025, and is eagerly awaited. There will be 63 issues to be determined at trial, and it will likely take a long time (several months at least) for that judgment to be handed down. The sheer size of that case shows the difficulty for copyright holders taking legal action in respect of the training of AI models. 

The interaction between copyright and AI is plainly a difficult issue to deal with, both legally and politically. This Kat doubts that the Lords' amendments to the Bill will make it on to the statute book, and expect that the Commons will decide that the outcome of the Copyright and AI Consultation ought to come first before any copyright reform. However, with the opt-out proposal losing support, and no proposal for an amendment to the list of restricted acts to address the jurisdictional challenges, transparency requirements appear to be the most achievable solution.