OpenAI revised the multi-agent orchestration in its Codex command line interface to encrypt messages passed to subagents, a change introduced via a pull request merged last month. The encryption applies to multi-agent v2 message payloads, which are the text instructions exchanged between agents. The pull request states that under the new system, "Responses encrypts the message argument returned by the model, Codex forwards only that ciphertext, and Responses decrypts it internally for the recipient model." OpenAI has not provided a reason for the change. Developers, including Ignat Remizov, CTO at payment service Zolvat, have expressed concern that the encrypted delivery path "removes the human-readable task/message text from local rollout history, trace reduction, and parent-side audit/debug surfaces," making debugging and auditing harder. Some developers speculate the encryption is intended to keep competitors from seeing how OpenAI's multi-agent implementation works. OpenAI did not immediately respond to a request for comment.