My Lords, I, too, will be relatively brief. I thank the noble Baroness, Lady Kidron, for her amendments, to which I was very pleased to add my name. She raised an important point about the practice of web scrapers, who take data from a variety of sources to construct large language models without the knowledge or permission of web owners and data subjects. This is a huge issue that should have been a much more central focus of the Bill. Like the noble Baroness, I am sorry that the Government did not see fit to use the Bill to bring in some controls on this increasingly prevalent practice, because that would have been a more constructive use of our time than debating the many unnecessary changes that we have been debating so far.
As the noble Baroness said, large language models are built on capturing text, data and images from infinite sources without the permission of the original creator of the material. As she also said, it is making a mockery of our existing data rights. It raises issues around copyright and intellectual property, and around personal information that is provided for one purpose and commandeered by web scrapers for another. That process often happens in the shadows, whereby the owner of the information finds out only much later that their content has been repurposed.
What is worse is that the application of AI means that material provided in good faith can be distorted or corrupted by the bots scraping the internet. The current generation of LLMs are notorious for hallucinations in which good quality research or journalistic copy is misrepresented or misquoted in its new incarnation. There are also numerous examples of bias creeping into the LLM output, which includes personal data. As the noble Baroness rightly said, the casual scraping of children’s images and data is undermining the very essence of our existing data protection legislation.
It is welcome that the Information Commissioner has intervened on this. He argued that LLMs should be compliant with the Data Protection Act and should evidence how they are complying with their legal obligations. This includes individuals being able to exercise their information rights. Currently, we are a long way from that being a reality and a practice. This is about enforcement as much as giving guidance.
I am pleased that the noble Baroness tabled these amendments. They raise important issues about individuals giving prior permission for their data to be used unless there is an easily accessible opt-out mechanism. I would like to know what the Minister thinks about
all this. Does he think that the current legislation is sufficient to regulate the rise of LLMs? If it is not, what are the Government doing to address the increasingly widespread concerns about the legitimacy of web scraping? Have the Government considered using the Bill to introduce additional powers to protect against the misuse of personal and creative output?
In the meantime, does the Minister accept the amendments in the name of the noble Baroness, Lady Kidron? As we have said, they are only a small part of a much bigger problem, but they are a helpful initiative to build in some basic protections in the use of personal data. This is a real challenge to the Government to step up to the mark and be seen to address these important issues. I hope the Minister will say that he is happy to work with the noble Baroness and others to take these issues forward. We would be doing a good service to data citizens around the country if we did so.