# Codex Now Encrypts Sub-Agent Prompts

Codex has implemented encryption for prompts exchanged between AI agents, preventing interception and tampering in multi-agent systems. This update addresses security concerns like prompt injection attacks, making agent-to-agent communications more resilient. The change is significant for enterprises relying on orchestrated AI workflows.

**Importance:** 3/5

## Sources

### Technology
- [Hacker News](https://github.com/openai/codex/issues/28058) — Tue, 14 Jul 2026 11:21:32 +0000