Security · 1h ago
OpenAI Codex Now Encrypts Prompts for Inference
OpenAI's Codex has started encrypting user prompts and performing inference on ciphertext, enhancing privacy. The change, noted in a GitHub issue, means prompts are encrypted before processing. This move aims to protect sensitive code and data during AI-assisted development.
Meridian48 take
While encryption during inference is a privacy win, the real test will be whether it impacts latency or accuracy in practice.
Read the full reporting
Codex starts encrypting prompts, uses ciphertext for inference instead →
Hacker News
openai-codexencrypted-inference