Daniel Hoadley, Head of Data Science and Analytics, Mishcon de Reya
We’re here tonight to talk about AI and IP law, so we have all these new technologies that are moving really, really quickly, they’re giving us incredible new capabilities but there are lots of tensions between these technologies and established law on IP.
Harry Clark, Associate, Mishcon de Reya
Welcome everyone, thank you all so much for coming. Today’s session on generative AI and IP: Securing protection and navigating the risks. So my name’s Harry, I’ll be chairing today’s panel alongside my lovely speakers, Nina, Peter and Anne.
Anne Rose, Managing Associate, Mishcon de Reya
So next we are going to come on as well and have a look at some of the practical tips and things that you really need to be thinking about. So first is whether AI generated work is actually protected by intellectual property rights for instance, copy rights. Another is actually looking at the data it’s been trained on, the materials, all the data that’s been fed into the AI tool and whether that’s infringing intellectual property rights and then finally you also really need to look at whether the AI generated image, the output is infringing third party intellectual property rights.
Nina O’Sullivan, Partner and Head Knowledge Lawyer, Mishcon de Reya
It relies on a lot of data, billions of pieces of data. How is that data collected? Is it licensed or is it scraped from the web? Where is it collected from, is it the internet generally or is it from a particular repository?
Harry Clark, Associate, Mishcon de Reya
So I think there’s two characteristics which has really driven a lot of interest that we’ve seen in terms of AI and that’s simply the pace and the scale of the technical developments. Within the last year or so a lot of these services are really leading into sort of customisation and privatisation so users can sort of tailor their experiences and tailor these models to apply to specific use cases and obviously if you are a business that creates some really interesting questions as to what it could mean for your current business practices and where they are going forward.
Peter Nunn, Partner, Mishcon de Reya
Put an image in to stable diffusion and ask it to create synthetic image of that image. What you get back is likely to be substantial reproduction of the image you just put into it and therefore be an infringement. Fine if you, if the image was your own and you own the copyright of it but not if you have taken someone else’s copyright work and asked for that image to be created.
Anne Rose, Managing Associate, Mishcon de Reya
The key as well, we touched a little bit on this earlier, we were looking at AI assisted works and AI generated works and I think it is really important that we actually make this distinction. You’ve got to give identification of the creator, there’s lots and lots of different conditions that you have to follow and comply with. So first off we really understand the technology and in order to do a proper analysis we have to understand the difference between various types of generative algorithmic models. So for instance, the difference between a diffusion model and an auto regressive model or even a generative adversarial network.
Daniel Hoadley, Head of Data Science and Analytics, Mishcon de Reya
We’re really well placed to talk about these things because not only do we have a fantastic IP team at Mishcon but we also have a fantastic data science team so there is this great confluence and understanding between technology, the law and this growing intersection between the two.
Anne Rose, Managing Associate, Mishcon de Reya
AI generated works pose intriguing questions about ownership, accountability and the evolving landscape with intellectual property rights and the technology driven era. We are really excited to see what happens next.