Tools · The Decoder ·

OpenAI's Codex now encrypts instructions between AI agents, leaving developers blind to internal delegation

OpenAI's Codex now encrypts instructions between AI agents, leaving developers blind to internal delegation

Since early June, OpenAI’s Codex has encrypted instructions that its main agent sends to subagents, preventing developers from tracking internal delegation. Encryption is mandatory for the GPT-5.6 variants Sol and Terra.

Read the full story at The Decoder →