Short paper: Models in the dark -- Rectification and erasure under GDPR in ML supply chains
A new paper explores the difficulties in enforcing GDPR's rights to rectification and erasure within machine learning systems. It highlights that current research often addresses these rights from either a legal or technical standpoint in isolation, neglecting the complex supply chains involved in ML model development and deployment. The paper introduces the concept of 'models in the dark'—downstream derived models lacking transparency—and analyzes the challenges they pose to data privacy. AI
IMPACT Highlights significant hurdles in applying data privacy regulations to complex ML systems, potentially impacting AI development and deployment strategies.