DOOMBENCH.COM
Safety and alignment

METR red team finds bypasses in OpenAI’s historical agent-monitoring stack

A privileged human red team found gaps in OpenAI’s July agent monitoring, including ways to delay review, conceal recent actions and potentially bypass sandbox controls. OpenAI says production vulnerabilities were patched and coverage expanded; METR has not verified those changes. This was authorized testing, not an autonomous escape.

0 comments · 0 votesOpen discussion

Public discussion is readable by everyone. Sign in to comment, reply, or vote.

No comments yet. Start the discussion.

Send detailed or private feedback →
CURRENT ASSESSMENT · REVISION 1
TOWARD DOOM57confidence 84/100

Why it moved the index

The exercise exposed weaknesses in monitoring used for real agent traffic, weakening evidence that deployment controls reliably detect dangerous behavior. Extensive internal access and partly unvalidated attack paths limit generalization. Historical coverage gaps do not prove a current compromise, and the vendor’s claimed remediation has not been independently checked.

Assessment history

  1. R1
    Toward 57 · confidence 84

    New dated primary disclosure of a distinct historical monitoring assessment, separated from prior agent incidents and autonomous-escape claims.

    24 Oct 2026

Share this page

DoomBench social sharing card for METR red team finds bypasses in OpenAI’s historical agent-monitoring stack.

  1. DoomBench assesses “METR red team finds bypasses in OpenAI’s historical agent-monitoring stack” as evidence moving toward doom, with magnitude 57 and confidence 84 out of 100 in the safety and alignment category.

  2. The DoomBench assessment of “METR red team finds bypasses in OpenAI’s historical agent-monitoring stack” is based on reporting from METR and records the editorial rationale, source quality, attribution, and revision history.

  3. DoomBench summarizes “METR red team finds bypasses in OpenAI’s historical agent-monitoring stack” as follows: A privileged human red team found gaps in OpenAI’s July agent monitoring, including ways to delay review, conceal recent...

    https://www.doombench.com/news/metr-red-team-finds-bypasses-in-openai-s-historical-agent-monitoring-stack-2026-10-22