-
Source: Stanford Institute for Human-Centered AI (HAI)
-
Released: 2026-04-13
-
Fetched: 2026-04-14
-
PDF: https://hai.stanford.edu/assets/files/ai_index_report_2026.pdf
-
Public data: https://drive.google.com/drive/folders/1zJTOg0iR0j5SijCwFutwWvDt143lW277
-
Local:
research/05-analyst-firms/raw/stanford-hai-ai-index-2026.pdf(37MB) -
Credibility: HIGH — named academic institution, multi-year longitudinal methodology, raw data published.
-
US–China model performance gap narrowed to 2.7 points on Arena (Claude Opus 4.6 1,503 vs ByteDance Dola-Seed Preview 1,464, March 2026).
-
SWE-bench Verified: 60% → near-100% in one year.
-
Generative AI reached 53% population adoption in three years — faster than PC or internet.
-
Organizational adoption 88%; 4 in 5 university students use generative AI.
-
Estimated consumer value $172B/yr in early 2026; median per-user value tripled 2025→2026.
-
Public–expert divide: only 10% of Americans more excited than concerned; 56% of AI experts expect positive 20-year impact.
-
Documented AI incidents: 233 (2024) → 362 (2025).
-
AI researcher inflow to US dropped 80% in the last year.
Next steps
- Extract chapter sections into topical notes under relevant pillars (01 landscape, 07 adoption, 09 adoption-cycle, 11 education, 06 security-frontier).
- Cross-reference banned-stat policy — do not quote any 95%/NANDA/MIT-2025/420% figures even if Stanford cites them indirectly.