AI Index

Confidence 0.85 · 2 sources · last confirmed 2026-04-30

The AI Index is an independent initiative at Stanford HAI that produces an annual report on the state of artificial intelligence. Conceived in 2017 within the One Hundred Year Study on AI (AI100). The 2026 edition is the 9th annual.

Mission (per the report): equip policymakers, journalists, executives, researchers, and the public with accurate, rigorously validated, globally sourced data on AI’s development, adoption, and impact.

What it tracks

Chapter (2026)Mapped to wiki concept
1. Research and Development (publications, patents, models, compute, energy)foundation-models
2. Technical Performance (benchmarks)ai-benchmarks
3. Responsible AI (incidents, transparency, governance)responsible-ai
4. Economy (adoption, investment, jobs, robots)enterprise-ai-adoption, ai-employment-effects
5. Science (new standalone in 2026)no concept page yet
6. Medicine (new standalone in 2026, with Schmidt Sciences)no concept page yet
7. Educationno concept page yet
8. Policy and Governanceno concept page yet
9. Public Opinionno concept page yet

In the 2025 edition, Science and Medicine were combined into a single chapter; the 2026 edition split them, adding a new ninth chapter overall.

Editions

YearEditionStatus
2017–20241–7Not yet ingested
20258Ingested — adds AI hardware deep-dive, novel inference cost estimates, fresh corporate adoption of responsible-ai data, expanded science/medicine coverage
20269Ingested — Medicine chapter spun off as standalone (with Schmidt Sciences); AI sovereignty as new analytical framework; new GenAI consumer-value estimates; “jagged frontier” enters the report’s narrative

(Earlier editions to be filled in as / if they’re ingested.)

Steering Committee (2026 edition)

Changes from 2025 edition

Data partners (2026 edition)

Center for Research on Foundation Models, CSTA, Digital Policy Alert, Epoch AI, ECEP (Expanding Computing Education Pathways), GitHub, International Federation of Robotics (IFR), Kapor Foundation, Lightcast, LinkedIn, McKinsey & Company, RAISE Health, Schmidt Sciences (new), Zeki.

Supporting funders (2026 edition)

Google, NSF, OpenAI, Open Philanthropy, Quid, Infosys (new).

Methodological notes

  • The report uses external surveys (notably McKinsey & Company’s Global AI Survey for adoption data) rather than original surveys for most adoption and use-case data. When citing AI Index adoption numbers, the underlying instrument is typically McKinsey’s, not a Stanford instrument.
  • AI tooling acknowledgement (2026): ChatGPT and Claude used to refine and copy-edit drafts; all images generated with AI (Johanna Friedman 2026, Gemini 3.1 / Gemini 3). The 2025 edition disclosed Claude/ChatGPT for copy-editing only.
  • License: CC BY-ND 4.0 (no derivatives). Public data and chart files released alongside the report on Google Drive; Global AI Vibrancy Tool covers 36 countries (up from ~30 in 2025).
  • Notable-models data partner: Epoch AI. Year-over-year counts may differ as new/historic models are added retroactively.

Open questions

  • Year-over-year stability of methodology (e.g., adoption numbers depend on McKinsey instrument changes).
  • The Global AI Vibrancy Tool (now 36 countries) — is it useful enough to file as its own resource page once engaged with?
  • The Medicine chapter being split off in 2026 is an editorial signal; comparison of 2025-combined vs 2026-split coverage of medicine may surface new findings worth ingesting.