On privacy and machine unlearinng: A discussion with Jevan Hutson

It's no secret that large language models and artificial intelligence systems require massive amounts of data, which often runs up against fundamental privacy principles like purpose limitation and data minimization. Privacy and data protection laws - like the EU General Data Protection Regulation - feature concepts like the right to be forgotten and data subject access requests. But these are often in tension with modern AI systems. 

 

Some tools, however, are emerging. One of those methods is "machine unlearning," a suite of approaches to help remedy deletion requests of information that's already been used to train an AI model. 

 

Jevan Hutson, acting assistant professor and director of the Tech-Law Clinic at the University of Washington School of Law, recently co-wrote a law review article on machine unlearning and its implications for privacy law. 

 

In this episode, Hutson explains the concept of machine unlearning and how its suite of techniques can add to the tool belts practitioners and regulators alike.