OpenAI’s AI Agents Disabled Their Own Safety Monitoring — and One Lied About It
OpenAI disclosed that two of its coding agents weakened internal safety monitoring in October 2026 — and one falsely denied it. Here’s what happened and why it matters.