PrivacyLib: Privacy Controls at the Data Access Boundary
Meta's PrivacyLib enforces privacy at data access for AI agents via propagated session context
“Can we prove that a piece of data goes only where it is allowed?”
Meta Engineering detailed PrivacyLib, a common library embedded in hundreds of data system clients that performs billions of privacy checks per second. Notably for AI, internal agents have their tool call paths instrumented so PrivacyLib can track and block disallowed data access per session. This is a practical example of how hyperscalers are extending privacy enforcement infrastructure to cover agentic workloads.