Stanford HAI (Institute for Human-Centered AI)

Stanford University’s cross-disciplinary AI institute. The wiki’s primary academic-grade source for AI adoption rates, capability benchmarks, agent evaluation, and policy data — the only annual academic-adjacent synthesis at this scope.

Notable outputs

• AI Index Report 2025 — hai.stanford.edu/ai-index/2025-ai-index-report — Annual report covering: business AI adoption (78% in 2025, up from 55% in 2023); capability benchmarks; agent eval (RE-Bench); investment trends; policy; education. Methodologically transparent; cross-year comparable.
• HELM benchmark — Holistic Evaluation of Language Models; broad benchmark suite covering accuracy, calibration, robustness, fairness, bias, toxicity, efficiency.

Why it matters for an enterprise wiki

Most analyst reports (Gartner, Insight, Knostic) publish point-in-time numbers without cross-year methodology. Stanford HAI’s AI Index is the academic-grade equivalent — annual, methodology-disclosed, cross-year comparable. For the wiki, it’s the canonical replacement for vendor-blog adoption-rate citations when triangulating claims like “75% knowledge-worker GenAI usage.”

See Also

• Source Triangulation Audit 2026-05-02 — primary citation for adoption-rate claims
• Agentic AI Threat Classes 2026 §Class 2 — capability scaling cross-reference