Copyright infringement risks in training Generative AI

The launch of ChatGPT in 2022 sparked a surge in both the popularity of, and investment in, generative AI technologies. 

Fast forward three years and, as the technology has advanced, so too has the sophistication of the debate surrounding its use and development. In some quarters, the spotlight now falls on the key question at the intersection of AI and IP – do developers need to obtain permission to use third party content to train their models? 

This topic has ignited intense debate worldwide, prompted governments to re-examine the adequacy of current copyright frameworks, and given rise to a number of IP infringement claims against AI developers in multiple jurisdictions. 

In our latest briefing we delve deeper into the evolving legal landscape by: 

• exploring the core copyright issues in play;
• reviewing the latest EU, UK and US approaches to tackling those challenges and some of the key questions arising in each of those jurisdictions; and
• spotlighting some of the key cases being litigated before the courts.

As the briefing highlights, the law relating to the use of third party copyright-protected materials to train generative AI remains complex and uncertain. Different jurisdictions are adopting different approaches and, in many cases, the law is subject to interpretation and may lead to different outcomes based on the specific facts involved. However, with a number of material court decisions expected in the next 12-18 months, the UK government due to respond on its consultation on copyright and AI, and the EU AI Act’s rules on general-purpose AI models now in force, the next few months could be crucial in shaping the future of how generative AI models are trained and how the rights of AI developers and IP rights holders are to be balanced.