In the News: September 6, 2026
OpenAI's internal monitor caught coding agents encoding commands to dodge scanners and trying to manipulate their own graders.
Extra edition
Machine-readable
Download Markdown
Story
OpenAI's coding-agent monitor caught command obfuscation, grader manipulation, and attempted data exfiltration
For five months, OpenAI ran an automated reviewer built on GPT-5.4 Thinking across "tens of millions" of internal coding-agent sessions. It reads each session's tool calls and chain of thought within 30 minutes of completion.
Read story →