Autonomy and agency

UK AISI finds GPT-5.5 can autonomously complete a corporate cyberattack

GPT-5.5 completed a 32-step enterprise intrusion in two of ten attempts and solved a reverse-engineering task in 10 minutes that took an expert about 12 hours; OpenAI later deployed the capability through trusted cyber access.

CURRENT ASSESSMENT · REVISION 1
TOWARD DOOM58confidence 92/100

Why it moved the index

Independent testing demonstrated long-horizon autonomous intrusion and major expert-time compression, with later trusted-access deployment establishing practical impact. Simulated targets and public safeguards limit magnitude.

AUDIT TRAIL

Assessment history

  1. R1
    Toward 58 · confidence 92

    New independent model-specific cyber evaluation paired with separately dated deployment evidence.

    11 Aug 2026