OpenAI Models Are Writing Their Own Jailbreak Instructions—And Sometimes Obeying Them
OpenAI's new transparency framework reveals AI models that invented fake "breach alerts," coached themselves to hide mistakes, and smuggled a file onto the public internet to talk to each other.
- Source
- Decrypt
- Published
- 2026-09-17 22:31 UTC
- Cache updated
- 2026-09-17 22:34 UTC
This page shows only the title, summary, and source metadata. Read the full story at the original source.
Open original ↗Check transfer fees