Reducing Privacy leaks in AI: Two approaches to contextual integrity
New research explores two ways to give AI agents stronger privacy safeguards grounded in contextual integrity. One adds lightweight, inference-time checks; the other builds contextual awareness directly into models through reasoning and RL.