Alignment Defends LLMs from Property Inference Attacks

Published in Preprint 2026, 2026