Copyright and Artificial Intelligence

Debate between Nusrat Ghani and Chi Onwurah

Wednesday 18th December 2024 Commons Chamber

[Madam Deputy Speaker (Ms Nusrat Ghani)]

I call the Chair of the Science, Innovation and Technology Committee.

[Chi Onwurah (Newcastle upon Tyne Central and West) (Lab)]

The UK is in a unique position—second in the world in the creative industries, and in the top three for AI innovation—so getting the right solution to protect and support our intellectual property, while supporting and incentivising AI innovation, is uniquely important to our cultural and economic life.

I am a former regulator and chartered engineer, so I welcome the Minister’s decision to go with regulatory technology as the solution, and to challenge the tech sector to come up with technology to ensure we can have both the reservation of rights and the transparency of inputs to large language models, both of which are critical.

The tech sector too often spends less time protecting people and property than maximising profit, but the language of the consultation is a bit vague. The Minister talked about arriving at a plan rather than a solution, so will he make it absolutely clear that any text and data mining exemption is contingent on the technology being deliverable, implementable and workable, and that if the technology fails, the exemption fails?

[Chris Bryant]

I welcome the Chair of the Select Committee to her place. She is 100% right that we cannot have the text and data mining exemption for commercial purposes unless there is a proper rights reservation system in place. I do not know whether she has looked at rights reservation, but it is terribly complicated. People can use the robots exclusion protocol, but it is rather out of date and is avoided by many players in the market. It is very complicated and applies only to a person’s own website, whereas their creative input might not be on their personal website—it might be on somebody else’s.

I tried to create a Bridget Riley using an AI bot over the weekend. The bot had obviously trained itself on some Bridget Riley works, but it was a shockingly bad Bridget Riley—it was nowhere near. I wanted to ask whether it had used Bridget Riley’s work to learn how to make a Bridget Riley-like picture and, if so, whether Bridget Riley received any compensation. Bridget Riley could use another website, haveibeentrained.com, if she wanted, but it is phenomenally complicated. That is precisely what must change. The AI companies must come up with a technical solution, whether they produce music, text or whatever. Without that, we will not be able to progress.