OpenAI's Codex CLI Multi-Agent V2 update has introduced encrypted inter-agent communication, removing developer visibility into instructions passed between parent and sub-agents. Developers on GitHub report they can no longer inspect task delegation, making debugging and root-cause analysis significantly harder. A proposed fix suggests keeping encrypted delivery while adding a separate plaintext audit field for traceability. Analysts warn that opaque agent orchestration creates serious governance, compliance, and auditability challenges for enterprises — particularly in regulated industries like banking and healthcare — and could slow production deployments of mission-critical AI systems.
612 Impressions