Report 002 · October 2026
AI and Extraordinary Human-Casualty Risk
Two independently generated October risk assessments, side-by-side model estimates, September comparisons, and the cumulative three-run assessment.
Independent research and comparisons of large-scale AI risk over successive assessments.
Independent research · Monthly archive
Independent five- and ten-year assessments of extraordinary AI-associated human-casualty risk, organized by research date. The October 2026 publication contains two separate model assessments and direct comparisons with September.
Two monthly publications, three research runs: September 2026 (GPT-5.6 Sol) and October 2026 (independent GPT-6 Astra Max and GPT-5.6 Sol assessments). The October page presents the models separately and compares both with September.
Report 002 · October 2026
Two independently generated October risk assessments, side-by-side model estimates, September comparisons, and the cumulative three-run assessment.
Report 001 · September 2026
Initial independently conducted five- and ten-year assessment, including forecasts, leading casualty pathways, attribution criteria, and research method.