00:00
2026-07-10
machinelearning.apple.com
ai-safety
Behavioral Privacy Leakage in Agentic Negotiation: Formalizing and Mitigating Inference Attacks via Randomized Policies
Researchers at the AI4TCI Workshop presented a method to protect behavioral privacy in autonomous negotiation agents, using randomized policies to reduce adversarial inference accuracy by 43–50% while…