Advanced Artificial Intelligence

Earl of Devon Excerpts

Monday 24th July 2023 Lords Chamber

[The Earl of Devon (CB)]

My Lords, having greedily signed up to all three debates taking place today because all are on topics close to my heart, I will try to keep the contributions short. I probably should have asked AI to write my speeches today—it would have saved time and doubtless made them much better. However, this is all my own work, and it is not wholly in AI’s favour.

I thank the noble Lord, Lord Ravensdale, for raising this issue. The development of advanced AI raises many significant risks and considerable opportunities, as with all technological advances. The technology itself is complicated, poorly understood and thus intimidating. As the noble Lord, Lord Ravensdale, noted, much of the public and policy debate is sensationalist as a result—we fear what we do not understand.

I do not pretend to understand the intricacies of AI engineering, so I will focus my attention on the implications for intellectual property rights. This is an area I know as an IP litigator qualified in the US and the UK, and I have clients in this space. I also note my interests as a member of the IP APPG, which champions the interests of IP rights holders. The APPG engaged with urgency last year following the Government’s announcement that they planned to introduce a new exception to copyright and database rights to permit text and data mining—or TDM—for any purpose. Currently TDM—in effect, the scraping of information from the internet—is permitted without a licence only for the purpose of academic research. The Government proposed to turbocharge the UK’s AI development by broadening this exception to allow TDM for all purposes, commercial or otherwise, irrespective of the rights and views of the authors of that material, in effect, riding a coach and horses through the long-established IP rights of individual creatives in the interests of the AI machines. The dystopian implications were clear.

Thankfully, sense and the APPG prevailed. The Government withdrew that proposal and have begun to investigate alternative solutions with the industry, such as collective licensing and consent that might respect authors’ rights. Can the Minister please give the House an update on those important discussions?

Despite this, since the public launch of multiple large language model AI systems, such as Open AI’s ChatGPT, Stability AI and others, it has become readily apparent that they have been extensively trained or “educated” using copyrighted materials for which no consent has been given. There is a real risk that AI is, in effect, larceny—an industry built on the theft of the personal property of creative authors for commercial ends. The flood of IP infringement cases that have followed would suggest this. I note, for example, Getty Images’ recent UK copyright infringement case against Stability AI’s image generation system and, more famously perhaps, Sarah Silverman’s American case against Open AI and Meta for theft of her written material.

Such is the brazen nature of this conduct that it seems that the creators of AI models consider they have no need to license protected works in training their machines—either because they are outside the jurisdiction or they consider that the “training” of their LLMs is educational. What steps have the Government taken to engage with them on this topic? I understand this is not the view of the Intellectual Property Office, which stated last year:

“Although factual data, trends and concepts are not protected by copyright, they are often embedded in copyright works. Data mining systems copy works to extract and analyse the data they contain. Unless permitted under licence or an exception, making such copies will constitute copyright infringement”.


Does the Minister agree with that statement and therefore that any large language model firms making copies without permission or exception will be infringing copyright?

Finally, can the Minister please update the House as to the Government’s thinking regarding AI inventions? When they reported last June, the Government had no plan to change the UK’s patent law because AI was not “yet” advanced enough to invent without human intervention. Given the rapid developments since, has the Government’s view changed on this point?

I note that the UK is co-ordinating closely with the United States on this issue. In the US in Thaler v Vidal, the Federal Circuit said that AI cannot be an inventor for US patents, but the USPTO has sought fresh evidence on the point. Similarly, the Copyright Office in the US rejected the registration of copyright in AI-generated works but has issued guidance and sought further evidence on the topic. What are the UK Government doing?

These are important issues. The sensitive balancing of the creative rights of the individual versus realising the extensive promise of machine learning is a key challenge for our times. Personally, I hope the humans will prevail.