Apple AI Updates: July 11, 2026

1. Apple Studies Privacy Leakage From How Negotiation Agents Behave

Apple. Apple researchers show that adversaries can infer a negotiation agent’s private constraints, such as reservation prices or deadlines, purely from behavioral signals like concession trajectories and response timing, even when the explicitly shared values are cryptographically protected. The paper formalizes these inference attacks and proposes randomized policies that inject controlled noise into agent behavior to mitigate the leakage. The work is relevant as more agent-to-agent transactions and automated negotiation move into production. Source