Misuse and incidents

Autonomous AI attack compromises Asian government entities

Dream's threat team recovered an attacker-operated autonomous system's working directory after twelve attack waves over four days. The campaign cracked 85 employee accounts, stole personnel and network data, and expanded toward government suppliers, a nuclear-safety agency, and energy companies. The underlying model and attacker remain unidentified.

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
TOWARD DOOM62confidence 68/100

Why it moved the index

Recovered operational artifacts and concrete external effects show autonomous cyber misuse scaling across real government and critical-infrastructure targets, directly increasing the demonstrated misuse pathway; confidence is limited because Dream has not published the underlying technical report and neither the attacker nor exact model is identified.

AUDIT TRAIL

Assessment history

  1. R1
    Toward 62 · confidence 68

    New dated primary incident account with concrete real-world effects, distinguished from the separately catalogued OpenAI and Hugging Face evaluation incident.

    13 Aug 2026