2026-W30
First retrospective baseline (week of 13–19 July 2026). The documents show active research on how easily AI safeguards can be bypassed and on automated red-teaming, plus an AI lab's view of US state and federal safety rules. No document reports a real-world loss of control or a serious incident this week. The index starts around the middle of the 'guarded' range with wide uncertainty.
A conventional composite index, not a scientific probability of extinction.