Why AI Can't Forget: The GDPR Problem with Large Language Models | S1 E8

13/08/2025 23 min Temporada 1 Episodio 8
Why AI Can't Forget: The GDPR Problem with Large Language Models | S1 E8

Listen "Why AI Can't Forget: The GDPR Problem with Large Language Models | S1 E8"

Episode Synopsis

In this episode of the Privacy Impact Shop brought to you by blackfootuk.com and gdpr.ie, Ben and Shaab are joined by PR technology expert Samantha Deeks to explore a fundamental challenge emerging at the intersection of AI and privacy law. The conversation centres around Samantha's revealing experiment with Claude AI, where she discovered the large language model had incorrect information about her company's co-founder—and more troubling still, admitted it couldn't "unlearn" this misinformation even when corrected. This discovery exposes a core architectural problem: neural networks, designed to replicate human thought processes, cannot selectively forget information the way traditional databases can delete records. This creates what may be an insurmountable compliance challenge with GDPR's right to rectification and erasure. The discussion examines how this "whisper campaign" effect—where AI provides personalised, unverifiable information in private conversations—could fundamentally change how reputational management works.Ben, Shaab, and Samantha explore the broader implications for the regulatory landscape, comparing the EU AI Act's use-case focused approach with traditional data protection frameworks. They discuss whether privacy laws will need to adapt to accommodate AI's technical limitations, or if AI regulation will need to become more prescriptive. The conversation also delves into the philosophical implications of AI regulation—how it's forcing us to redefine human values and what behaviours we consider fundamentally unacceptable, regardless of legal compliance. The episode concludes with practical advice for PR professionals and business leaders about understanding AI's reputational risks, the importance of organisations proactively auditing what AI systems know about them, and why governance frameworks may be more critical than regulatory compliance in managing AI-related harms.Perfect for privacy professionals, PR practitioners, business leaders implementing AI strategies, and anyone grappling with how artificial intelligence is reshaping our relationship with information, truth, and personal data rights.