A new Tsinghua University paper reveals that AI agents with long-term memory fail to cleanly delete personal data upon user request. Testing across six memory systems showed an end-to-end success rate of less than 1.5%, with most deleted facts still extractable via simple prompts. The study urges developers to implement and verify actual deletion paths rather than relying on superficial confirmation messages.
Termsslug: machine-unlearning
Machine Unlearning
机器遗忘
Techniques allowing models to remove the influence of specific training data without full retraining.
Curated Intelligence
Rohan Paul (@rohanpaul_ai)T2·Agents
78 pts
Google ResearchT1·Models
68 pts
Researchers propose a new framework for auditing machine unlearning without access to model internals or original training data. By leveraging statistical methods like two-sample testing on output queries, it mathematically verifies privacy deletion, significantly reducing compliance costs and improving audit efficiency.