Jul 14, 2026 11:33 UTC
hacker news
https://news.ycombinator.com/item?id=48905028
Codex appears to be implementing prompt encryption and using ciphertext directly for inference.
READY TO POST:
rumors swirling that codex is now encrypting prompts and inferring from ciphertext. seems like a big privacy win if true.
codex might be encrypting prompts before inference. this could be a really interesting move for data privacy in ai applications.