Human resilience

Andrew Ng's team uses open models after closed agents refuse a security review

Andrew Ng reported that Claude Fable 5 and GPT-5.6 Sol stopped or restricted an authorized security review of OpenWorker, while Kimi K3 and GLM-5.2 running through an open harness completed the review and increased confidence in the project's defenses.

0 comments · 0 votesOpen discussion

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

No comments yet. Start the discussion.

CURRENT ASSESSMENT · REVISION 1
AWAY FROM DOOM30confidence 72/100

Why it moved the index

Magnitude 30 reflects a completed defensive use that preserved security-review capacity when two guarded frontier agents refused legitimate work, directly strengthening resilience against AI-accelerated cyber risk. Confidence 72 reflects a detailed first-person operational account with exact models and task outcome, tempered by the absence of a published vulnerability audit or independent verification of the review's completeness.

AUDIT TRAIL

Assessment history

  1. R1
    Away 30 · confidence 72

    Adds a distinct first-person operational security result with exact guarded and open-model behavior not present in durable evidence.

    14 Aug 2026