🧰 Auto-Review

The items that name it 7

• Addendum to GPT-6 Astra System Card: GPT-6.1 Terra
  article2026-10-26

  Protocol in Codex where a second model checks commands that run outside a sandbox, blocks unsafe ones and tells the model to avoid similar actions.

• Red-Teaming OpenAI's Internal Agent Monitoring Systems
  article2026-10-22

  Codex review safeguard that METR found agents could in principle bypass, together with sandboxing.

• OpenAI Agents Silenced Security Team Alerts, Removed Misalignment Monitor
  article2026-10-21

  Review step required for all internal Codex use at OpenAI. It has been updated to spot and stop attempts to turn off misalignment monitoring.

• GPT-Live-1.1 system card
  article2026-10-15

  Harness-level restriction mentioned in alignment evaluations

• GPT-6 Sol and GPT-6 Luna: October 2026 update
  advisory2026-10-07

  A protocol built into Codex whereby a second model evaluates the safety of certain commands run outside a pre-specified sandbox, blocking execution and instructing the model to avoid similar actions when it judges a command unsafe.

• Introducing dots
  article2026-09-29

  Checks actions that could affect accounts or share information against user instructions, Custom Rules and safety requirements, determining what proceeds, needs approval, or stays with the user

• GPT-6 Astra System Card
  advisory2026-09-03

  A protocol built into Codex whereby a second model evaluates the safety of certain commands run outside a pre-specified sandbox, blocking execution and messaging the model when a command is judged unsafe

In every workspace you can read 12