[PODCAST] Data Protection and Privacy in the Age of Artificial Intelligence: In Conversation with Mr Adrian Mak

‘What is the legitimate interest when billions of people did not really say [they had a] reasonable expectation that [their personal data] would be used to train AI in the first place?’

When using artificial intelligence, we often care more about what the answer is rather than where it comes from. However, as Adrian Mak, a Fellow at the Stanford Law School AI Initiative explains, we should think more about the fact that the answer is usually generated from the sensitive personal data used to train such models. 

A co-editor of ‘Privacy and Personal Data Protection Law in Asia’ by Hart Publishing and a contributor to ‘The Cambridge Handbook of Private Law and Artificial Intelligence’, Mak has extensive practice experience in international technology, commercial and energy disputes. In Hilary Term 2026, Isaac Tan, a Podcast Editor of the Oxford University Undergraduate Law Journal, sat down with Mak to discuss the underlying norms of data protection and privacy and how these norms may be challenged and maintained in the Age of AI. 

Mak explains that there are two main ways in which our personal data is threatened in the age of AI. The first issue is that Large Language Models (LLMs) rely on training models such as Common Crawl or LAION-5B, which contain the personal data of billions of individuals. The question, according to Mak, then becomes whether consent has been given in processing the personal data of those individuals. Even if there has been consent, there may still be residual questions of legitimate interest in processing the data. The second issue is one of output. Mak argues that there is a real danger that AI models, when pushed hard enough, will simply regurgitate the highly sensitive personal data they were trained with, as demonstrated in the ongoing litigation of Getty Images v Stability AI and New York Times v Stability AI. 

Certain legal and technical solutions have emerged in response to these threats, which, as Mak explains, are complex and often involve competing considerations of profitability and innovation. The recent Draghi Report argues that the legal solutions adopted in the West run a risk of stifling AI innovation. Mak, however, has a different view. Citing the aviation sector as an example, Mak argues that increased regulation and calls for safety may actually boost improvements in AI technology. 

The discussion concludes with a discussion on the competing notions of replacement and augmentation. Mak explains that replacement theory sees AI replacing humans in the workforce while augmentation theory sees AI augmenting the way humans work. As a proponent of augmentation theory, Mak encourages everyone, from law students to legal practitioners, to engage and play around with different models. Far from seeing AI as competition and a replacement for work, Mak argues that we should be thinking about how AI can augment and enhance our workflows.  

‘I think [norms are] essential, it’s no longer just a good to have nowadays…Within the AI circle, people always talk about replacement rather than the augmentation of human labour…and the current vibe shift in Silicon Valley and Stanford is to do this alignment work…and it’s the intersection of [ethical and practical] that is the most interesting.’ 

The Spotify link to the episode is available here.

A transcript of the episode is available here.