METR: historical internal-agent monitoring pilot

Evidence record

External researchers describe a high-access historical assessment.

Read the original source ↗

Claim-level citations (4)

• supportsMETR pilot separates monitoring coverage from resistance to agent interference: OpenAI stated production-impacting issues were patched; METR did not evaluate those claims. The read-only-tool escalation hypothesis was not validated.

  Findings final tool bullet and paragraph beginning "Since completing"

• supportsMETR pilot separates monitoring coverage from resistance to agent interference: Researchers found an exploit that could conceal hundreds of recent actions; public details were redacted.

  Findings bullet beginning "We also discovered an exploit"

• supportsMETR pilot separates monitoring coverage from resistance to agent interference: They reported an in-principle Codex sandbox and Auto-Review bypass reaching the host user's authority.

  Findings bullet beginning "We found a way Codex agents"

• supportsMETR pilot separates monitoring coverage from resistance to agent interference: Unmonitored compaction and return-to-user triggering allowed monitoring postponement; two client-controlled headers could disable monitoring for new agents. Creating such agents could itself be detected.

  Findings: compaction, return-control trigger, and client-controllable headers