Comprehensive annual survey of artificial intelligence progress, applications, and societal impact in 2025, covering research and development, technical performance, responsible AI, economy, science, medicine, education, policy and governance, and global public opinion.
Quick Navigation
Tags
Share the slides
Comprehensive annual survey of artificial intelligence progress, applications, and societal impact in 2025, covering research and development, technical performance, responsible AI, economy, science, medicine, education, policy and governance, and global public opinion.
Detailed view of each slide page, including layout, key content and visual elements.
Full cover of the Artificial Intelligence Index Report 2026, 9th edition, featuring Stanford HAI branding with a red/coral accent and abstract geometric pattern.
Full-bleed cover with centered title, edition number, and Stanford HAI logo at bottom
Formal title page listing Stanford Human-Centered AI (HAI) as publisher, editorial team credits, copyright notice, and publication details for the 9th annual edition.
Clean white page with centered text hierarchy, HAI logo top, publication metadata bottom
First page of the table of contents listing chapters 1 through 5 with page numbers, chapter titles, and brief one-line descriptions of each chapter's scope.
Two-column layout with chapter numbers on left, titles and page references on right; color-coded chapter tabs
Second page of the table of contents covering chapters 6 through 9 plus appendix, with page numbers and one-line chapter descriptions.
Continuation of two-column TOC layout matching page 3 style
Opening foreword from HAI directors including Fei-Fei Li, introducing the significance of the 9th annual AI Index and contextualizing AI's rapid advancement in 2025.
Single-column text with director photo inset top-right, HAI header banner
Continuation of the foreword discussing AI's transformative societal impact, the importance of data-driven AI policy, and HAI's mission to guide human-centered AI development.
Single-column text layout with section subheading
Third foreword page covering HAI's approach to tracking AI progress across multiple dimensions and acknowledging the global team of contributors to the 2026 report.
Single-column text with acknowledgment paragraph and contributor credits
Closing page of the foreword with final remarks from HAI leadership on the stakes of AI development in 2025-2026 and a call to action for policymakers and researchers.
Single-column text closing with director signatures block
Opening of the executive summary presenting the overarching narrative of AI in 2025: rapid capability gains, accelerating investment, growing policy activity, and societal debates.
Full-width intro paragraph followed by numbered list preview of 10 key takeaways
Key takeaways 1 through 3: (1) AI capabilities surging with benchmark saturation; (2) US private AI investment hits $285.9B; (3) AI incidents and harms tracking shows 1000% increase since 2019.
Three-panel layout each with takeaway number, bold headline, supporting data point, and small icon
Key takeaways 4 through 6: (4) AI agents now resolve 50%+ of real GitHub issues; (5) FDA approved 258 AI-enabled medical devices in 2025; (6) 46 countries have adopted national AI strategies.
Three-panel layout matching page 10 style
Key takeaways 7 through 9: (7) 80% of students use generative AI for coursework; (8) EU AI Act prohibited practices ban took effect February 2025; (9) 59% of global respondents optimistic about AI.
Three-panel layout matching pages 10-11
Final executive summary takeaway (10): AI Nobel Prize recognition marks historic milestone for science, plus a highlights chart showing key metrics across all nine chapters at a glance.
Single takeaway panel top-half; summary metrics dashboard bottom-half
Visual summary chart displaying year-over-year growth metrics across major report dimensions: investment, publications, model releases, policy actions, and adoption rates.
Full-width grouped bar chart with color-coded categories and year labels on x-axis
World map visualization showing AI activity levels by country across dimensions of research output, private investment, policy adoption, and talent concentration.
Full-width world map with color gradient intensity scale and country callouts for top nations
Brief explanation of the AI Index's data collection methodology, defining what constitutes AI in the report's scope, and listing primary data sources used across chapters.
Two-column text with source list in sidebar and methodology diagram inset
Navigation guide for readers explaining chapter structure, how to find specific topics, online interactive data access, and citation guidelines for the AI Index 2026.
Single-column text with call-out boxes for online resources and chapter navigation tips
Full-bleed chapter opener page with dark navy/slate grey background, large white Chapter 1 numeral, and title 'Research and Development' with a brief one-paragraph chapter overview.
Full-bleed dark background, oversized chapter number left-aligned, title and description right-aligned
Introductory text for Chapter 1 summarizing key R&D findings: approximately 200 notable AI models released in 2025, benchmark saturation accelerating, open-source models closing the gap with proprietary systems.
Single-column introduction with highlighted key statistics pull quotes
Bar chart showing the number of notable AI model releases per year from 2019 to 2025, illustrating the dramatic acceleration with approximately 200 notable models released in 2025 alone.
Full-width bar chart with year labels on x-axis, count on y-axis, bars colored in chapter navy/slate theme
Stacked bar chart breaking down 2025 AI model releases by organization type: industry, academia, government, and civil society, showing industry's dominant share.
Stacked horizontal bar chart with organization type categories and legend
Geographic breakdown of AI model origins in 2025 showing US, China, UK, Europe, and other regions, with the US producing the largest share of notable frontier models.
World map with bubble sizes indicating model release volume per country, plus accompanying bar chart
Overview section on foundation models in 2025, defining the term, listing key frontier models (GPT-4o, Gemini 2.0, Claude 3.5, Llama 3, DeepSeek), and explaining their architectural significance.
Text-heavy section intro with model name callout boxes and a comparison table
Horizontal timeline visualization showing the release dates of major foundation models throughout 2025, organized by organization and model family.
Horizontal timeline with model labels above/below the line, organization color-coding
Line chart comparing performance trajectories of open-source versus proprietary AI models from 2020-2025 on standardized benchmarks, showing the narrowing gap by late 2025.
Dual-line chart with open-source (dashed) and proprietary (solid) lines, benchmark score on y-axis, year on x-axis
Analysis of the open-source AI ecosystem growth in 2025, including GitHub star counts, download volumes for top models, and the role of platforms like Hugging Face in democratizing AI access.
Mixed layout with text analysis left and bar charts right showing download/star metrics
Line chart showing total AI-related academic publications from 2010-2025 by region, demonstrating China's lead in volume and the US's continued leadership in highly-cited papers.
Multi-line chart with regional color coding, log-scale y-axis option, year on x-axis
Scatter plot of AI publication volume versus citation impact by country, illustrating the US-China publication volume-quality dynamic and the EU's citation performance.
Scatter plot with country bubbles sized by total publication count, x-axis volume, y-axis citations per paper
Stacked area chart showing AI publication growth by research subdomain (NLP, computer vision, robotics, ML theory, etc.) from 2015-2025, with NLP and generative AI showing explosive growth.
Stacked area chart with subdomain color layers, year x-axis, publication count y-axis
Bar chart comparing AI research publication venues — top conferences (NeurIPS, ICML, CVPR, ICLR, ACL) versus journals — and their relative share of high-impact AI research in 2025.
Side-by-side bar chart comparison with venue names on y-axis, paper count on x-axis
Introductory analysis of benchmark saturation in AI, explaining how many previously challenging benchmarks have been largely solved by 2025, necessitating harder evaluation tasks.
Text section with benchmark saturation diagram showing tasks moving from 'unsolved' to 'saturated' zones
Scatter plot showing AI performance versus human performance across 50+ benchmarks, with color-coding indicating benchmarks where AI has surpassed human-level performance.
Large scatter plot with human-parity diagonal line, benchmark name annotations, color-coding for above/below human parity
Log-scale line chart tracking training compute (in FLOPs) used for frontier AI models from 2012-2025, showing the approximately 6-month doubling time and orders-of-magnitude increases.
Log-scale line chart with model name annotations, time on x-axis, log FLOPs on y-axis
Analysis of AI training cost trends, showing that while compute per model increases, the cost per unit of compute has fallen dramatically, with a dual-axis chart illustrating both trends.
Dual-axis chart with compute cost per FLOP on left axis and training cost per model on right axis
Charts showing improvements in AI inference efficiency from 2020-2025: tokens per second, cost per million tokens, and model size required for equivalent performance all improving dramatically.
Three-panel small multiple charts each showing one efficiency metric over time
Overview of AI hardware developments in 2025: GPU/TPU advances, new custom AI accelerators from major tech firms, and the emergence of inference-optimized chips.
Text overview with comparative hardware specs table and market share pie chart
Bar chart and text analysis showing growth in training dataset sizes used for frontier models, from billions to trillions of tokens, and the shift toward synthetic data generation.
Bar chart showing dataset size by model with text commentary column
Analysis of multimodal AI capabilities in 2025 — models integrating text, image, audio, and video — with capability matrix showing which models support which modalities.
Capability matrix table with model names as rows, modalities as columns, checkmarks/scores in cells
Section on the emergence of reasoning-focused models in 2025 (e.g., OpenAI o-series, DeepSeek-R1), explaining chain-of-thought at inference time and benchmark improvements on hard reasoning tasks.
Text section with before/after comparison chart showing reasoning model vs standard model performance
Introduction to AI agent systems released in 2025, covering architectures that enable multi-step autonomous task completion, tool use, and agentic workflows.
Conceptual diagram of agent architecture (perception → reasoning → action loop) with text explanation
Comparative analysis of AI R&D leadership by country across dimensions: publication volume, citation impact, patent filings, model releases, and private investment.
Radar/spider chart per top country (US, China, UK, EU, Canada, India) with five-dimension axes
Bar chart showing AI patent filings by country from 2015-2025, with China surpassing the US in raw patent volume while the US maintains leadership in highly-cited patents.
Grouped bar chart by country with year groupings, patent count on y-axis
Data on new AI company formations in 2025, showing the surge in AI-native startups and the geographic distribution of new AI ventures across the US, EU, and Asia.
World map bubble chart for startup formation plus bar chart of top startup hubs
Analysis of market concentration in frontier AI model development, showing that a small number of organizations (OpenAI, Google, Anthropic, Meta, Mistral, DeepSeek) account for most frontier releases.
Pie chart showing share of frontier model releases by organization plus text analysis
Analysis of AI research talent flows: graduate program enrollment trends, geographic migration of AI researchers, and the concentration of top AI talent in a few institutions and companies.
Sankey diagram showing talent flows between countries plus bar chart of top institutions
Ranked list and bar chart of top AI research institutions by publication output and citation impact in 2025, including universities and industry labs.
Ranked bar chart horizontal with institution names on y-axis, dual metrics on x-axis
Timeline of the most significant AI research and development milestones in 2025, from model launches to scientific breakthroughs and policy events.
Vertical timeline with milestone entries, dates on left, descriptions on right, category color-coding
Overview of AI progress in robotics and embodied AI in 2025: foundation models for manipulation, locomotion improvements, and the emergence of generalist robot policies.
Text overview with robot capability benchmark chart and illustrative photos
Analysis of generative AI progress for image and video synthesis in 2025, including diffusion models, video generation quality improvements, and real-time generation capabilities.
Text analysis with quality metric progression chart and example capability descriptions
Overview of generative AI advances in audio and speech synthesis in 2025: voice cloning, music generation, real-time translation, and ambient speech recognition improvements.
Text overview with performance metric charts for speech recognition error rates and voice quality scores
Section on AI code generation advances in 2025, covering model capabilities on HumanEval and SWE-bench, and the integration of AI coding assistants across developer workflows.
Performance benchmark progression chart plus text analysis with developer adoption statistics
Analysis of emerging AI research themes gaining traction in 2025: mechanistic interpretability, AI alignment, continual learning, test-time compute scaling, and neuromorphic approaches.
Topic cluster visualization with publication growth charts per theme
Data on international AI research collaboration measured by co-authorship across borders, showing patterns of collaboration and increasing US-China research decoupling.
Chord diagram of international co-authorship flows with text commentary
Breakdown of AI research funding sources in 2025: government grants, industry funding, philanthropic investment, and the growing share of industry-funded academic research.
Stacked bar chart over time showing funding source shares, with text analysis
Analysis of disparities in AI compute access between large tech firms, academic institutions, and researchers in lower-income countries, highlighting the compute divide.
Comparative bar chart showing compute allocation with text analysis on access inequality
Data on AI's growing energy consumption in 2025: estimated energy use of frontier model training, inference at scale, and data center power demand growth attributed to AI workloads.
Line chart of AI-attributed energy demand with carbon equivalence annotations and text commentary
Discussion of major open research problems in AI as of 2025: long-horizon reasoning, reliable planning, robustness to distribution shift, interpretability, and compositional generalization.
Structured text with problem-statement boxes and citation counts per research challenge
Summary statistics page for Chapter 1 presenting the most important quantitative findings in a dashboard format: model releases, compute growth, publication volumes, and open-source metrics.
Dashboard grid of large-number statistics with supporting sparklines and brief text labels
Reference notes for Chapter 1 citing primary data sources, methodology notes for publication and patent data, and definitions of key terms used in the R&D chapter.
Two-column footnote/endnote format with source citations and numbered references
Curated list of further reading resources and key papers referenced in Chapter 1, organized by section topic, with brief annotations.
Structured reading list with section headers and annotated entries
Full-bleed chapter opener with dark navy/blue background, large white Chapter 2 numeral, title 'Technical Performance', and brief overview of AI performance advances measured in 2025.
Full-bleed dark navy background, oversized chapter number, title and description in white
Introduction to the technical performance chapter, explaining the benchmarking framework, caveats about benchmark saturation, and the organization of performance data across task categories.
Single-column introduction text with chapter section overview diagram
Overview of language understanding benchmark performance in 2025 across MMLU, MMLU-Pro, BIG-Bench Hard, and HellaSwag, showing AI models at or above human performance on most tasks.
Multi-bar comparison chart with benchmark names on y-axis and performance scores on x-axis for top models
Line chart tracking top model performance on MMLU and MMLU-Pro benchmarks from 2020-2025, demonstrating rapid progression from 60% to near-human and superhuman levels.
Dual-line chart with MMLU and MMLU-Pro trends, model milestone annotations, human performance reference
Bar chart showing AI performance on competition mathematics benchmarks (AIME, AMC, MATH) in 2025, with leading reasoning models achieving top human-competitive scores.
Grouped bar chart by benchmark with model comparison bars, human expert baseline references
Line chart tracking AI performance on MATH dataset from 2020-2025, showing the inflection point in 2024-2025 when reasoning models dramatically improved mathematical problem-solving.
Line chart with key model annotations, inflection point highlighted, human score reference line
Comparison of AI coding performance on HumanEval, HumanEval+, and MBPP benchmarks in 2025, showing near-saturation on original HumanEval and strong performance on harder variants.
Bar chart comparing top coding models across three benchmarks with human developer baseline
Line chart and bar chart showing AI agent performance on SWE-bench (real GitHub issue resolution) from 2023-2025, with leading agents now resolving over 50% of issues.
Left: line chart of best performance over time. Right: bar chart of top agents with resolution rate
Analysis of AI performance on agentic coding tasks requiring multi-file edits, test writing, debugging across large codebases, and tool use — going beyond single-function HumanEval style tasks.
Benchmark comparison table with task category rows and model performance columns
Bar chart showing top model performance on GPQA (Graduate-Level Google-Proof Q&A) across physics, chemistry, and biology subdomains, with frontier models surpassing PhD-level human performance.
Grouped bar chart by science domain with model comparison and PhD-level human baseline
Multi-benchmark comparison across scientific reasoning tasks (GPQA, SciQ, ARC-Challenge, OlympiadBench) showing AI model performance relative to expert human baselines.
Radar chart per top model showing multi-benchmark performance profile
Analysis of AI performance on formal logic, theorem proving, and structured reasoning tasks in 2025, including performance on LeanDojo, FrontierMath, and logical deduction benchmarks.
Bar chart of performance by reasoning task category with text analysis
Performance analysis of AI models on long-context tasks requiring comprehension of documents up to 1 million tokens, showing significant improvement but remaining challenges in very long contexts.
Line chart of performance versus context length with model comparison lines
Analysis of AI model factual accuracy and hallucination rates across 2023-2025, showing improvements in grounded generation but persistent hallucination challenges in open-domain settings.
Bar chart comparing hallucination rates across models with improvement trend line
Data on AI model performance on instruction-following benchmarks (IFEval, MT-Bench) in 2025, showing strong gains in following complex, multi-constraint instructions.
Bar chart with instruction complexity tiers on x-axis and compliance rate on y-axis for top models
Benchmark results for vision-language models in 2025 across tasks like visual QA, image captioning, and chart understanding, showing AI approaching or exceeding human performance.
Multi-benchmark comparison bar chart for top vision-language models
Performance analysis of AI video understanding models on temporal reasoning, action recognition, and video QA benchmarks in 2025.
Bar chart comparing top video models on multiple benchmarks with human performance baseline
Word error rate trends for AI speech recognition systems from 2015-2025 across multiple languages and accent groups, showing near-human performance in high-resource languages.
Multi-line chart of WER over time by language group
Human evaluation scores for AI image generation quality from 2021-2025, showing dramatic improvements in photorealism, coherence, and prompt adherence.
Line chart of quality scores over time with annotated model releases and sample output descriptions
Bar chart showing AI agent task completion rates on standardized agentic benchmarks (WebArena, OSWorld, AgentBench) in 2025, highlighting progress toward general autonomous task execution.
Grouped bar chart by benchmark with agent system comparison bars and baseline references
Analysis of AI agent performance on web browsing, computer use, and GUI interaction tasks, showing success rates and failure mode analysis for leading agent systems.
Bar chart of success rates by task type with failure mode breakdown chart
Data on AI agent performance on long-horizon planning tasks requiring 10+ sequential steps, showing current limitations and the gap between short-task and long-task performance.
Line chart of performance degradation with increasing task steps for multiple agent systems
Analysis of AI translation quality across 200+ language pairs using BLEU and human evaluation scores, showing near-professional quality for high-resource pairs and significant gaps for low-resource languages.
World language heat map for translation quality plus bar chart of top/bottom performing pairs
Benchmarks for AI text summarization and long-form generation quality in 2025, including human evaluation consistency scores and factual accuracy metrics.
Bar chart comparing summarization quality across models with human evaluation results
Analysis of AI model robustness to distribution shift, adversarial inputs, and out-of-distribution test cases, showing persistent fragility despite overall capability gains.
Grouped bar chart comparing in-distribution vs OOD performance gaps across model types
Consolidated comparison chart showing AI versus human performance across 30+ distinct tasks, with visual markers for tasks where AI now exceeds human performance.
Horizontal dot plot with task names on y-axis, AI and human scores as paired dots, 'AI exceeds' shading
Performance data on newly introduced hard benchmarks designed to resist saturation in 2025: FrontierMath, MMMU-Pro, and Humanity's Last Exam, showing current frontier model capabilities.
Bar chart of top model scores on three hard benchmarks with human expert baseline
Analysis of the test-time compute scaling paradigm in 2025: how additional inference compute through chain-of-thought, search, and self-verification improves performance on hard tasks.
Line chart showing performance vs inference compute budget for reasoning models
Analysis of AI model calibration — how well model confidence scores match actual accuracy — showing improvements in calibration for top models but remaining challenges.
Calibration curve plots for multiple models with ideal calibration diagonal reference
Heat map showing AI model performance across 100+ languages on standard NLP benchmarks, highlighting the significant performance gap between high-resource and low-resource language groups.
Language-by-benchmark heat map with performance color gradient and regional groupings
Benchmark performance summary across specialized professional domains: law (Bar exam), medicine (USMLE), finance (CFA), and accounting, showing frontier AI models at or above professional passing scores.
Bar chart of professional exam pass scores vs AI model scores across four domains
Overview of AI performance in strategic games and simulations in 2025, including chess, Go, poker, and real-time strategy games, showing superhuman performance across most game categories.
Rating comparison chart for AI vs top human players across game types
Benchmark performance data for embodied AI systems on manipulation, navigation, and dexterous control tasks, showing substantial progress but remaining gap to human performance.
Bar chart of success rates by task category for top robotic AI systems
Analysis of AI performance per dollar of inference cost in 2025, showing dramatic improvements in cost efficiency that are democratizing access to high-capability AI.
Scatter plot of performance vs cost per query for top models with time-based trend arrows
Head-to-head capability comparison of the top 10 frontier models of 2025 across 15 benchmark categories, providing a structured overview of the current AI capability landscape.
Large comparison table with model names as rows, benchmark categories as columns, color-coded performance tiers
Summary line charts showing performance trends across five major task categories (language, math, code, science, multimodal) from 2018-2025, illustrating the acceleration of AI progress.
Five-panel small multiple line charts each showing one category trend over time
Methodological discussion of AI benchmark design limitations, including data contamination concerns, benchmark overfitting, and the challenge of measuring general intelligence.
Text analysis with illustrative contamination diagram and checklist of benchmark quality criteria
Data on AI model performance on safety and alignment-relevant benchmarks: TruthfulQA, BBQ bias evaluation, and jailbreak resistance testing, showing progress and remaining challenges.
Bar chart comparing safety benchmark scores across top models with text commentary
Reference notes for Chapter 2 with data source citations, benchmark methodology references, and definitions of performance metrics used throughout the chapter.
Two-column endnote format with numbered citations and source references
Curated further reading list for Chapter 2, organized by benchmark category, with key papers and leaderboard resources for continued exploration of AI technical performance.
Structured reading list with category headers and annotated entries
Deep-dive analysis of chain-of-thought prompting effects on reasoning performance, showing that explicit step-by-step reasoning improves performance across math, logic, and science benchmarks.
Bar chart comparing with/without CoT performance plus text analysis of mechanism
Analysis of retrieval-augmented generation (RAG) system performance in 2025, showing how combining LLMs with external knowledge retrieval improves factual accuracy significantly.
Bar chart of factual accuracy with and without RAG across knowledge domains
Benchmark performance of AI models on structured data tasks: SQL generation, table QA, and financial data interpretation, showing strong but imperfect performance.
Bar chart of structured data task performance with task type breakdown
Analysis of reproducibility challenges in AI performance benchmarking, including sensitivity to prompt phrasing, temperature settings, and evaluation protocol variations.
Variance analysis chart showing performance sensitivity to prompt variation with text commentary
Discussion of emergent AI capabilities observed in 2025 — skills that appear discontinuously as model scale increases — with examples and debate about whether emergence is real or artifact of evaluation.
Text analysis with emergence examples table and scale versus capability scatter plot
Overview of the AI benchmarking ecosystem in 2025: key leaderboards, evaluation organizations (HELM, BIG-bench), and the challenge of maintaining meaningful evaluation as models improve.
Ecosystem diagram showing major benchmarks, organizations, and their relationships
Performance data on multimodal reasoning tasks requiring integration of text and visual information, including science diagrams, charts, and math with figures.
Benchmark comparison bar chart for top multimodal models across reasoning task types
Analysis of the gap between benchmark performance and real-world deployment performance, including discussion of distribution shift, edge cases, and the benchmark-deployment performance correlation.
Scatter plot comparing lab benchmark score to real-world performance estimates for applications
Chapter 2 summary dashboard presenting the headline performance findings: AI at human parity on X tasks, surpassing human experts on Y tasks, and key outstanding challenges.
Summary dashboard with large-number statistics, task category breakdown, and key-finding callout boxes
AI performance on competitive programming benchmarks (CodeForces, IOI problems) in 2025, showing frontier models reaching high-Codeforces-rated performance levels.
Rating comparison chart with AI system levels mapped to human percentile equivalents
Detailed analysis of AI performance on International Mathematical Olympiad (IMO) style problems in 2025, with frontier models achieving gold-medal-equivalent performance on selected problem sets.
Problem difficulty vs solution rate chart with model comparison and IMO medal threshold annotations
Performance of AI systems on scientific literature review tasks: accurately summarizing papers, identifying methodology errors, and synthesizing findings across multiple papers.
Accuracy comparison bar chart for literature review sub-tasks with model comparison
Detailed multilingual benchmark performance (XTREME, XGLUE, mMMLU) showing performance gaps between English and non-English tasks, with analysis of low-resource language challenges.
Heat map of performance by language and benchmark with language family groupings
Research findings on AI persuasion capabilities in 2025: studies showing AI-generated persuasive text outperforming human-written text in A/B tests, raising concerns about manipulation.
Bar chart of persuasion study results with text analysis of implications and risk discussion
Analysis of AI model memory capabilities in 2025: in-context memory limits, external memory augmentation, and stateful agent systems enabling longer-horizon tasks.
Comparative diagram of memory architectures with performance data on memory-dependent tasks
Performance data on AI models using external tools (APIs, calculators, search engines, code interpreters) in 2025, showing substantial improvements in reliable tool use.
Success rate bar chart for tool use tasks by tool type and model
Overview of domain-specific AI models for scientific tasks: protein structure, materials science, climate, genomics, and their performance relative to general-purpose models.
Performance comparison table: specialist vs generalist models across scientific domains
Benchmark performance for AI document understanding tasks: PDF parsing, form extraction, contract analysis, and complex document QA in 2025.
Bar chart of document task performance with task type breakdown and model comparison
Analysis of AI model robustness against adversarial attacks in 2025, including image perturbation attacks, prompt injection for LLMs, and defense mechanism effectiveness.
Attack success rate bar chart with defense method comparison and text analysis
Concluding analysis of Chapter 2 key themes: the acceleration of AI capability gains, the challenge of keeping benchmarks meaningful, and the implications of AI surpassing human performance on an expanding set of tasks.
Text conclusion with summary chart showing benchmark categories by human-parity status
Full-bleed chapter opener with dark teal background, large white Chapter 3 numeral, title 'Responsible AI', and brief overview of the chapter's focus on AI safety, bias, harms, and governance.
Full-bleed dark teal background, oversized chapter number, title and description in white
Introduction to the responsible AI chapter explaining the framework for measuring AI harms, safety research, bias evaluation, and governance in practice across 2025.
Single-column introduction with section overview diagram and key definitions box
Line chart showing the number of AI-related incidents tracked by the AIAAIC database from 2012-2025, demonstrating more than 1000% growth over the period with steep acceleration from 2022.
Full-width line chart with year x-axis, incident count y-axis, key incident spikes annotated
Stacked bar chart breaking down AI incidents in 2025 by harm category: disinformation/deepfakes, privacy violations, discrimination, autonomous system failures, and cybersecurity breaches.
Stacked horizontal bar chart with harm category color coding and percentage labels
World map and bar chart showing the geographic distribution of reported AI incidents in 2025, with the US, EU, and China accounting for the majority of tracked incidents.
World map with incident density bubbles plus bar chart of top 15 countries by incident count
Analysis of AI-generated disinformation incidents in 2025: deepfake videos targeting elections, synthetic audio fraud, and AI-generated news articles spread at scale.
Timeline of major 2025 deepfake incidents with text analysis and detection capability chart
Comparison of deepfake detection technology performance in 2025 versus deepfake generation quality, showing the ongoing arms race with detection consistently lagging generation.
Dual-line chart of generation quality vs detection accuracy over time with arms-race gap annotation
Overview of bias measurement and mitigation research in AI systems in 2025, covering hiring algorithms, credit scoring, healthcare triage, and facial recognition across demographic groups.
Framework diagram of bias types and mitigation approaches with text overview
Experimental results from bias audits of AI hiring tools in 2025, showing differential screening rates by race, gender, and age in automated resume screening systems.
Bar chart of screening rates by demographic group with disparity ratio annotations
Analysis of bias in healthcare AI algorithms in 2025, focusing on differential accuracy of diagnostic AI across racial and socioeconomic groups and the causes of performance gaps.
Bar chart of diagnostic accuracy by demographic group across multiple AI tools
Data on AI bias in credit scoring and financial services, showing differential loan approval rates and risk score distributions across demographic groups in AI-assisted lending.
Bar chart of approval rates and score distributions by demographic with text analysis
Benchmark results for facial recognition systems across demographic groups in 2025, showing continued accuracy disparities despite progress, with analysis of regulatory responses.
Error rate comparison chart by demographic group with regulatory response callouts
Line chart tracking the volume of AI fairness and bias mitigation research publications from 2015-2025, showing the field's rapid growth alongside increasing corporate fairness commitments.
Line chart of publication volume plus bar chart of corporate fairness initiative announcements by year
Line chart showing the growth of AI safety research publications and dedicated safety research organizations from 2016-2025, demonstrating the field's rapid institutionalization.
Dual-line chart of safety publications and safety org count over time
Overview of major AI safety research organizations active in 2025: Anthropic, DeepMind Safety, ARC Evals, Redwood Research, METR, Apollo Research, and others.
Organization profile cards in a grid with founding date, focus area, and key outputs per org
Analysis of AI red-teaming adoption across major AI labs in 2025, showing red-teaming has become standard practice, with emerging standards from NIST and government frameworks.
Adoption timeline chart plus text analysis of red-teaming methodologies and findings
Summary of disclosed red-teaming findings across major AI labs in 2025: common vulnerability categories, jailbreak technique prevalence, and the efficacy of different defense approaches.
Vulnerability category bar chart with defense efficacy comparison table
Overview of AI-generated content watermarking and detection technologies in 2025: C2PA provenance standards, AI text detectors, and image watermarking adoption by major platforms.
Technology landscape diagram with adoption metrics and text analysis of effectiveness
Analysis of AI privacy risks and protective measures in 2025: training data memorization, model inversion attacks, differential privacy adoption, and regulatory compliance challenges.
Risk taxonomy diagram with adoption rate chart for privacy-preserving techniques
Analysis of AI's dual role in cybersecurity in 2025: AI-powered threat detection and defense versus AI-enabled attack automation, with data on incident rates for both.
Side-by-side bar charts for AI defense applications and AI attack applications with text analysis
Overview of capability evaluation frameworks used by frontier AI labs in 2025 to assess dangerous capabilities before model deployment, including CBRN uplift testing.
Evaluation framework diagram with capability category definitions and lab adoption table
Comparison of major AI risk taxonomies and frameworks in 2025: NIST AI RMF, EU AI Act risk tiers, OECD AI risk framework, and industry-developed risk classifications.
Side-by-side comparison table of four frameworks with risk categories mapped across frameworks
Data on NIST AI RMF adoption rates across US federal agencies and private sector organizations in 2025, showing growing but uneven uptake.
Adoption rate bar chart by sector and organization size with text analysis
Analysis of model card and AI transparency documentation adoption in 2025, showing increased disclosure by major labs but inconsistent standards and depth across the industry.
Adoption metrics bar chart plus quality scorecard comparing model card depth across major labs
Overview of AI applications in military and defense contexts in 2025: autonomous weapons development, targeting AI, AI-enabled surveillance, and international law challenges.
Text analysis with global deployment map and risk factor callout boxes
Data on AI use in criminal justice settings in 2025: predictive policing, sentencing algorithms, bail decision tools, and documented bias concerns and legal challenges.
Adoption map by US state with bias audit results chart and legal challenge timeline
Global analysis of AI-powered surveillance technology deployment in 2025: facial recognition in public spaces, behavior monitoring, and country-level adoption and restrictions.
World map of surveillance technology adoption with regulatory restriction overlay
Analysis of growing AI environmental impact concerns in 2025, focusing on data center water usage, carbon emissions from AI workloads, and sustainability initiatives by major AI companies.
Bar chart of environmental metrics per company with sustainability commitment comparison table
Overview of labor concerns related to AI in 2025: data labeling worker conditions, content moderation worker mental health, and AI-driven job displacement discussions.
Text analysis with survey data on labeler working conditions and job displacement projection charts
Analysis of AI's role in content moderation on major platforms in 2025, including detection accuracy for harmful content categories and the challenge of false positives for marginalized communities.
Accuracy metrics bar chart by content type with false positive rate breakdown by community
Overview of corporate AI responsible use policies at major AI companies in 2025, including usage policies, prohibited use cases, and enforcement mechanisms.
Policy comparison table across major AI companies with use-case prohibition checklist
Survey data on AI governance practices in industry: percentage of firms with AI ethics boards, responsible AI teams, impact assessments, and governance review processes.
Bar chart of governance practice adoption rates by firm size and sector
Overview of international AI safety coordination efforts in 2025: the Paris AI Safety Summit, the Bletchley Park follow-up, the Seoul AI Declaration, and multilateral safety research initiatives.
Timeline of international safety summits and agreements with text analysis of commitments
Three detailed case studies of significant AI incidents in 2025: a major deepfake election disinformation campaign, a healthcare AI diagnostic error cluster, and an autonomous vehicle accident series.
Three-panel case study layout with incident description, response, and lessons learned per case
Data on AI model performance on dangerous capability evaluation benchmarks in 2025, including CBRN uplift assessments and biosecurity risk evaluations used in pre-deployment testing.
Bar chart of capability evaluation results with risk tier thresholds and lab comparison
Overview of AI alignment research approaches in 2025: RLHF/RLAIF, Constitutional AI, debate and amplification, mechanistic interpretability, and scalable oversight techniques.
Research approach taxonomy diagram with publication volume chart per approach
Analysis of AI interpretability and explainability research growth in 2025, covering mechanistic interpretability techniques, saliency maps, and their practical deployment in regulated industries.
Publication growth line chart plus adoption rate in regulated industries bar chart
Data on AI mental health applications and associated risks in 2025: chatbot-based mental health support deployment, efficacy data, and incidents involving harmful AI mental health interactions.
Deployment metrics with efficacy chart and incident category breakdown
Overview of child safety concerns related to AI systems in 2025: CSAM detection, age-appropriate AI experiences, children's AI use in education, and regulatory protections.
Text analysis with regulatory landscape map and safety measure adoption chart
Experimental data on the effectiveness of leading bias mitigation techniques in 2025, comparing pre-processing, in-processing, and post-processing approaches across bias metrics.
Comparative bar chart of bias metric reduction by technique with accuracy cost analysis
Data on responsible AI job roles in 2025: AI ethicist, AI safety engineer, responsible AI program manager postings, salaries, and the supply-demand gap in the field.
Job posting trend chart and salary comparison with supply-demand gap callout
Overview of the emerging AI auditing ecosystem in 2025: third-party auditors, algorithmic audit standards, mandatory audit requirements in various jurisdictions, and industry self-assessment tools.
Ecosystem diagram of AI audit organizations plus regulatory requirement comparison table
Analysis of responsible AI concerns specific to open-source model releases in 2025, including dual-use risks, post-release safety, and community safety practices.
Risk taxonomy for open-source AI with mitigation approach comparison and text analysis
Summary of outcomes from the Paris AI Safety Summit (February 2025) and subsequent international safety commitments, including the GPAI Code of Practice milestones.
Outcome summary cards per summit with commitment tracking status indicators
Summary dashboard for Chapter 3 presenting key responsible AI metrics: AI incident growth, bias audit results, safety research growth, and governance adoption rates.
Dashboard grid with large-number statistics, trend sparklines, and key finding callout boxes
Analysis of AI system tendencies to generate toxic, harmful, or inappropriate outputs in 2025, including toxicity benchmark results and the effectiveness of safety fine-tuning.
Toxicity rate comparison bars with model improvement trends and text analysis
Overview of AI copyright and intellectual property challenges in 2025: training data rights, AI-generated content IP status, major legal cases, and emerging regulatory frameworks.
Legal case timeline with policy framework comparison and text analysis
Analysis of documented AI-enabled disinformation campaigns targeting elections in 2025, including scale estimates, platform responses, and detection challenges.
Campaign timeline with scale estimates and platform response comparison table
Analysis of accountability gaps in AI deployment chains in 2025: diffuse responsibility across developers, deployers, and users, and efforts to assign liability.
Accountability gap diagram with liability framework comparison across jurisdictions
Overview of AI systems' impact on human rights in 2025, from surveillance and free expression to access to justice and the right to explanation in automated decisions.
Rights impact matrix table with AI application rows and human rights columns, with impact rating cells
Preview of technical safety evaluation methodologies referenced in Chapter 3, including red-team protocols, capability evaluation procedures, and bias measurement standards.
Methodology overview diagram with reference list and text explanations
World map visualization of responsible AI regulatory and governance activity in 2025, showing regulation density by country and the contrast between EU, US, and Asian approaches.
World choropleth map with regulation intensity color coding and regional callout boxes
Comparative analysis of AI ethics principles adopted by governments, companies, and international organizations in 2025, mapping overlaps and divergences.
Venn/overlap diagram of ethics principle themes across 20+ major frameworks
Reference notes for Chapter 3 including primary data sources for incident data, bias studies, safety research, and governance framework citations.
Two-column numbered endnote format
Curated further reading for Chapter 3 organized by section: safety research, bias and fairness, governance, and incident analysis, with brief annotations.
Structured reading list with section headers and entries
Full-bleed chapter opener with deep navy blue background, large white Chapter 4 numeral, title 'Economy', and brief chapter description covering AI investment, labor, and adoption trends.
Full-bleed deep navy background, oversized chapter number, white title and description
Introduction to the economy chapter, framing AI's economic footprint in 2025 across private investment, labor markets, firm adoption, and productivity effects.
Single-column introduction with chapter highlights callout boxes and section overview
Line chart showing global private AI investment from 2013-2025, reaching over $470 billion globally in 2025, with the US accounting for $285.9 billion, a tripling from 2022 levels.
Line chart with global and US-specific investment lines, year x-axis, dollar amount y-axis, key milestone annotations
Bar chart ranking countries by private AI investment in 2025, with the US leading significantly, followed by China, UK, India, and Canada.
Horizontal ranked bar chart with country names on y-axis and investment amounts on x-axis
Breakdown of AI private investment by industry sector in 2025: enterprise software, healthcare, autonomous vehicles, financial services, cybersecurity, education, and retail.
Pie chart and accompanying bar chart showing sector investment shares and amounts
Analysis of the generative AI investment surge in 2025, with a bar chart showing generative AI's share of total AI investment growing from 15% in 2023 to over 45% in 2025.
Bar chart of generative AI investment share by year with total investment overlay line
Data on AI company funding mega-rounds ($100M+) in 2025 and the global AI unicorn count, showing acceleration in billion-dollar valuations for AI-native companies.
Bubble chart of mega-round deals by company with table of top 20 AI unicorns
Analysis of AI-related mergers and acquisitions activity in 2025, showing record M&A volumes and the strategic consolidation of AI talent and technology by large technology companies.
M&A deal volume bar chart by year with deal type breakdown and key deal callouts
World map and Gini coefficient analysis showing the geographic concentration of AI private investment, with the US, China, and UK accounting for over 80% of global AI investment.
World choropleth map of investment concentration plus concentration metrics panel
Analysis of Fortune 500 companies' AI adoption in 2025: over 90% mention AI in earnings calls, with McKinsey survey data showing 70% using AI in at least one business function.
Adoption trend line chart plus sector breakdown bar chart with mentions-to-adoption funnel diagram
Bar chart comparing AI adoption rates across firm sizes (large, mid-size, small, micro) showing significant adoption gaps and the challenges facing smaller firms in accessing AI tools.
Grouped bar chart by firm size with adoption metric breakdown and text analysis
Sectoral breakdown of AI adoption in 2025: financial services and technology leading adoption, followed by healthcare, retail, and manufacturing, with education and government trailing.
Horizontal bar chart ranking sectors by AI adoption rate with technology, use-case, and investment data
Summary of empirical studies measuring AI productivity effects across industries in 2025, showing documented gains in coding productivity, knowledge work output, and customer service efficiency.
Study results summary table with productivity gain estimates and text analysis of methodology
Analysis of AI productivity gains in knowledge work tasks: writing, coding, research, analysis, and customer service, with field study estimates ranging from 20-80% efficiency gains.
Bar chart of productivity gains by task type with study source citations and error bars
Line chart showing AI-related job postings growth from 2019-2025, with approximately 30% year-over-year growth in 2025 and AI skills appearing in a growing share of non-AI-specialist postings.
Dual-line chart of AI specialist postings and AI-skill-required postings over time
Bar chart showing wage premiums for AI-skilled workers in 2025, with AI roles commanding 30-40% wage premiums over comparable non-AI roles across software engineering, data science, and research.
Wage premium bar chart by role type with comparison to non-AI equivalents
Analysis of occupational groups' exposure to AI automation and augmentation in 2025, showing high-wage cognitive work highly exposed to augmentation and routine task workers at displacement risk.
Scatter plot of occupational wage versus AI exposure with occupation cluster labels
Analysis of macro-level employment data in AI-leading economies, showing strong employment in AI-intensive sectors despite automation concerns, with nuanced sectoral patterns.
Employment trend charts by sector with AI-intensity overlay and text analysis
Analysis of AI's distributional effects on wages: evidence of wage gains concentrated in high-education, high-skill workers, with limited wage growth for low-skill workers.
Wage growth comparison chart by education level with distribution curve insets
Summary of current empirical evidence on AI-related job displacement in 2025, distinguishing documented cases from projected displacement scenarios, with sector-level displacement data.
Displacement evidence table with sector rows, documented vs projected columns, and text commentary
Analysis of the AI skills gap in 2025: shortage of AI-skilled workers across virtually all sectors, the gap between AI job demand and qualified supply, and retraining challenges.
Supply-demand gap chart by role and region with text analysis of gap causes
Discussion of AI's potential role in exacerbating economic inequality between high-AI-capability nations and lower-income countries, and between high-skill and low-skill workers within economies.
Text analysis with inequality projection chart and regional capability gap visualization
Data on AI infrastructure investment in 2025: data center construction, GPU/chip procurement, electricity grid investments, and hyperscaler capital expenditure driven by AI demand.
Infrastructure investment bar chart by category with year-over-year growth annotations
Analysis of AI's impact on financial markets in 2025: AI company stock performance, AI-driven trading algorithm concerns, and investor sentiment data for AI sector stocks.
AI sector stock index performance chart with market cap callouts and text analysis
Data on AI deployment in customer service in 2025: chatbot adoption, resolution rate improvements, cost reduction metrics, and consumer satisfaction with AI customer service.
Adoption metrics chart with satisfaction comparison and cost reduction analysis
Analysis of AI adoption in marketing and advertising in 2025: generative AI for content creation, AI-driven ad optimization, personalization at scale, and ROI evidence.
Adoption rate chart with ROI evidence table and content creation cost reduction data
Overview of AI adoption in legal services in 2025: contract review automation, legal research AI, e-discovery tools, and the impact on legal work hours and billable rates.
Time savings bar chart by legal task type with adoption rate and text analysis
Data on AI applications in manufacturing and supply chain in 2025: predictive maintenance, quality control vision systems, logistics optimization, and productivity gains.
ROI and efficiency gains chart by application type with adoption rate breakdown
Analysis of AI deployment in financial services in 2025: fraud detection, credit underwriting, algorithmic trading, compliance, and the regulatory landscape for financial AI.
Use case adoption chart with fraud reduction metrics and regulatory callout boxes
Survey data on AI adoption among small and medium enterprises in 2025, showing growing adoption of AI tools for marketing, customer service, and operations, with access barriers.
Adoption rate trend with barrier analysis chart and SME sector breakdown
Data on the AI startup ecosystem in 2025: new company formation rates, sector focus areas, funding rounds, and success metrics for AI-native versus AI-enabled startups.
Startup ecosystem diagram with funding stage funnel and sector focus pie chart
Comparison of national AI economic development strategies in 2025: US AI investment, EU AI industrial policy, China AI economy plan, India AI mission, and others.
Comparative strategy table with investment amounts and priority areas per country
Analysis of AI's role in international trade in 2025: AI in customs and logistics, AI export controls, chip trade restrictions, and AI's impact on comparative advantage.
Export control timeline with trade flow impact chart and text analysis
Summary of major economic projections for AI's economic impact by 2030, from McKinsey, Goldman Sachs, and IMF, showing consensus on significant but uncertain GDP impact.
Projection comparison chart with scenario ranges and methodology notes
Analysis of AI-enabled new business models in 2025: AI-as-a-service, vertical AI solutions, autonomous agent services, and platform businesses built on foundation model APIs.
Business model taxonomy diagram with example companies and revenue model descriptions
Data on AI applications in the energy sector in 2025: grid optimization, renewable energy management, demand forecasting, and oil and gas exploration AI use.
Application bar chart with efficiency gain metrics and sector deployment map
Overview of AI adoption in agriculture in 2025: precision farming, crop disease detection, yield prediction, and agricultural robotics deployment and productivity data.
Adoption trend chart with productivity gain data and geographic distribution map
Analysis of AI applications in real estate in 2025: property valuation algorithms, market prediction tools, construction AI, and smart building management systems.
Application adoption chart with valuation accuracy metrics and text analysis
Analysis of AI compute market economics in 2025: GPU rental prices, cloud AI service pricing trends, hyperscaler AI revenue, and the economics of training vs inference.
Price trend charts for compute with market size estimates and hyperscaler AI revenue comparison
Salary and compensation data for AI roles in 2025, including ML engineers, AI researchers, data scientists, and AI product managers across different company types and geographies.
Salary comparison chart by role and geography with total compensation breakdown
Analysis of the AI IP economy in 2025: AI-generated IP valuation, training data licensing markets, patent royalties for AI techniques, and legal uncertainty impacts on investment.
IP market size estimate chart with legal uncertainty timeline and text analysis
Regional AI economy spotlights covering US AI corridor cities, EU AI hubs, China's AI technology parks, India's AI ecosystem, and emerging AI economies in Southeast Asia.
Regional map with city/hub callout cards and key metrics per region
Chapter 4 summary dashboard with headline economic figures: $285.9B US AI investment, 90%+ Fortune 500 AI mention rate, 30-40% AI wage premium, and documented productivity gains.
Summary dashboard with large-number statistics and chapter color-themed graphics
Reference notes for Chapter 4 with investment data sources, labor market study citations, and productivity research methodology notes.
Two-column numbered endnote format
Curated further reading for Chapter 4 covering AI investment, labor economics, productivity research, and economic impact projections with brief annotations.
Structured reading list with section headers
Full-bleed chapter opener with dark teal/forest background, large white Chapter 5 numeral, title 'Science and AI', and brief overview of AI's transformative role in scientific research in 2025.
Full-bleed dark teal/forest background, oversized chapter number, white title and description
Introduction to Chapter 5 framing AI as a new tool for scientific discovery in 2025, with overview statistics on AI-assisted research growth and the AI Nobel Prize milestone.
Single-column introduction with key statistics callout boxes and section overview
Line chart showing the growth of AI-assisted science publications from 2015-2025 across major disciplines, with 71% growth in AI-assisted biology and life science research highlighted.
Multi-line chart by scientific domain with year x-axis and publication count y-axis
Analysis of AlphaFold 3 and subsequent protein structure prediction advances in 2025, with benchmark accuracy data and the growth of AI-assisted protein research.
Accuracy benchmark chart with structural prediction examples table and text analysis
Overview of AI prediction of complex biomolecular structures beyond proteins in 2025: protein-DNA, protein-ligand, RNA structure, and antibody design using foundation models.
Capability comparison chart with text analysis and key breakthrough callout boxes
Analysis of AI-discovered drug candidates in 2025: number entering clinical trials, disease areas targeted, cost reduction estimates (30-50% savings), and key AI-drug-discovery companies.
Drug pipeline funnel diagram with disease area breakdown and cost reduction bar chart
Data on AI-discovered drugs currently in clinical trials in 2025, including Phase I/II/III status, therapeutic areas, and the first AI-discovered drugs approaching regulatory approval.
Clinical trial status table with pipeline chart and milestone timeline
Overview of AI applications in materials science in 2025: AI-designed battery materials, semiconductors, catalysts, and structural materials, with examples entering production.
Materials application bar chart with discovery timeline and text analysis
Analysis of AI applications in genomics in 2025: variant effect prediction, gene regulation modeling, single-cell analysis, and AI-guided CRISPR targeting.
Application capability matrix with genomics AI timeline and text analysis
Overview of AI applications in climate and earth science in 2025: AI climate models showing 10,000x speedups, downscaling global models, extreme event prediction, and carbon cycle monitoring.
Speedup comparison chart with application domain breakdown and text analysis
Performance comparison of AI weather forecasting models (GraphCast, Pangu-Weather, FourCastNet) versus traditional NWP models in 2025, showing AI exceeding traditional models on most metrics.
Performance comparison bar chart for multiple forecast metrics with traditional model baseline
Analysis of AI earthquake prediction improvements in 2025: early warning time extensions, magnitude estimation accuracy, and aftershock prediction using deep learning on seismic data.
Performance improvement chart with geographic deployment map and text analysis
Overview of AI applications in physics research in 2025: particle physics data analysis at CERN, quantum system simulation, plasma physics for fusion, and AI-guided experimental design.
Application domain timeline with key discovery callouts and text analysis
Analysis of AI applications in astronomy in 2025: galaxy classification at scale, gravitational wave signal detection, exoplanet discovery from telescope data, and cosmological simulation acceleration.
Discovery count chart with application domain breakdown and text analysis
Overview of AI in chemistry research in 2025: reaction prediction, synthesis planning, catalyst design, and AI-assisted chemical safety screening, with publication growth data.
Publication growth chart with application breakdown and key tool callouts
Discussion of the historic 2024 Nobel Prize recognition of AI contributions to science (Geoffrey Hinton — Physics; Demis Hassabis and John Jumper — Chemistry), marking official scientific establishment recognition of AI.
Timeline of Nobel Prize announcements with laureate profiles and significance analysis
Data on AI tools processing scientific literature in 2025: semantic search over millions of papers, automated hypothesis generation, claim extraction, and citation graph analysis.
Tool capability comparison table with adoption rate chart and text analysis
Overview of AI systems conducting autonomous scientific experiments in 2025: robotic lab systems guided by AI, self-directed hypothesis generation, experiment design, and result interpretation.
Autonomous science pipeline diagram with capability milestone timeline and text analysis
Analysis of AI applications in neuroscience in 2025: brain mapping at scale, neural signal decoding, brain-computer interface improvements, and AI models of neural computation.
Application capability chart with brain mapping scale data and text analysis
Performance data for AI in biomedical imaging research (distinct from clinical use) in 2025: single-cell imaging analysis, cryo-EM processing, and super-resolution microscopy AI.
Performance benchmark chart by imaging modality with text analysis
Overview of AI applications in social science research in 2025: large-scale survey analysis, natural language processing of historical texts, and AI-assisted behavioral experiments.
Application type overview with publication growth chart and text analysis
Analysis of AI research infrastructure supporting scientific AI in 2025: national AI research clouds, open-access computing resources for science, and the role of NSF, DOE, and international equivalents.
Infrastructure map with resource comparison table and funding analysis
Analysis of AI tools for mathematical research in 2025: automated theorem proving, AI-assisted proof verification, conjecture generation, and the growth of AI-assisted mathematics publications.
Capability progress chart with theorem proving benchmark data and text analysis
Discussion of AI tools for improving scientific reproducibility in 2025: automated protocol documentation, code and data sharing tools, and AI-assisted result verification.
Reproducibility improvement diagram with adoption metric charts and text analysis
Analysis of AI tools in scientific peer review in 2025: automated review assistance, paper screening, ethical violation detection, and debates about AI reviewer reliability.
Adoption rate chart with reliability data and ethical debate callout boxes
Analysis of geopolitical dimensions of AI-enabled science in 2025: research collaboration patterns, AI science capability gaps between nations, and open-access AI science tools.
Collaboration network diagram with capability gap map and text analysis
Overview of science-specific foundation models in 2025: biology (ESM3, Geneformer), chemistry (ChemBERTa), materials (MatBERT), earth science (ClimaX), and cross-domain scientific LLMs.
Model taxonomy table with capability comparison and domain coverage diagram
Data on AI's impact on clinical trial design and execution in 2025: patient recruitment optimization, protocol design assistance, adverse event detection, and trial timeline reductions.
Timeline reduction chart with application breakdown and text analysis
Overview of AI applications in epidemiology and public health research in 2025: disease surveillance, outbreak prediction, genomic epidemiology, and AI-assisted vaccination strategy modeling.
Application capability chart with surveillance deployment map and text analysis
Analysis of AI contributions to energy science research in 2025: fusion plasma control with AI, battery discovery, carbon capture material design, and solar cell optimization.
Application timeline with performance improvement data and text analysis
Analysis of cross-disciplinary impact of AI science tools, showing how AI is enabling breakthroughs at the interfaces between biology, chemistry, physics, and engineering.
Cross-disciplinary impact diagram showing AI tool flows between scientific domains
Discussion of risks and challenges in AI-enabled science: AI hallucinations in scientific contexts, over-reliance on AI models, black-box scientific reasoning, and the deskilling of researchers.
Risk taxonomy with case study examples and mitigation discussion
Chapter 5 summary dashboard: key statistics on AI-assisted science publication growth, breakthrough milestones, and sectors where AI is having the greatest scientific impact.
Summary dashboard with large-number statistics and chapter teal-themed graphics
Reference notes for Chapter 5 with primary data sources for scientific publication data, AlphaFold citations, and domain-specific research references.
Two-column numbered endnote format
Curated further reading for Chapter 5 organized by scientific domain, including key papers and review articles on AI in biology, chemistry, earth science, and cross-domain AI science agents.
Structured reading list with domain section headers and annotated entries
Full-bleed chapter opener with deep blue background, large white Chapter 6 numeral, title 'Medicine and AI', and brief description of AI's growing role in clinical care and biomedical research.
Full-bleed deep blue background, oversized chapter number, white title and description
Introduction to Chapter 6 presenting AI in medicine as a transformative force in 2025, with key statistics: 258 FDA-approved AI devices, 85.5% diagnostic accuracy for MAI-DxO, 300K+ clinicians using AI scribes.
Single-column introduction with headline statistics callout boxes and section overview
Benchmark comparison of AI diagnostic systems versus specialist physicians in 2025 across radiology, pathology, dermatology, and ophthalmology, showing AI achieving or surpassing specialist accuracy.
Multi-domain accuracy comparison bar chart with specialist baseline and AI performance bars
Detailed analysis of MAI-DxO achieving 85.5% diagnostic accuracy on standardized clinical cases, surpassing the average specialist physician performance and validating AI diagnostic potential.
Accuracy comparison chart with case type breakdown and statistical significance analysis
Bar chart showing FDA-approved AI-enabled medical device approvals by year from 2016-2025, reaching 258 approvals in 2025 and totaling 950+ cumulative approvals.
Bar chart with year x-axis and annual approval count on y-axis, cumulative line overlay
Breakdown of FDA-approved AI medical devices by specialty in 2025: radiology dominating at ~75%, followed by cardiology, pathology, and other specialties.
Pie chart and bar chart of device distribution by medical specialty
Detailed analysis of AI in radiology in 2025: CT analysis, chest X-ray interpretation, mammography screening, and MRI analysis, with clinical deployment data and diagnostic accuracy benchmarks.
Performance benchmark chart by imaging modality with deployment statistics and text analysis
Overview of AI computational pathology in 2025: digital pathology slide analysis, cancer grading assistance, biomarker detection, and the growing clinical deployment of AI pathology tools.
Capability comparison chart with deployment data and text analysis
Analysis of AI applications in ophthalmology in 2025: diabetic retinopathy screening, glaucoma detection, age-related macular degeneration diagnosis, and retinal image analysis.
Diagnostic accuracy bar chart by condition with geographic deployment map and text analysis
Analysis of ambient AI clinical documentation scribes in 2025, used by 300,000+ clinicians to reduce documentation burden, with data on time savings and burnout reduction.
Adoption trend line with time savings metrics and burnout reduction survey data
Clinical outcome data from AI documentation scribe deployments in 2025: physician time savings (1-2 hours per shift), patient interaction time increases, and documentation accuracy improvements.
Before/after comparison charts for time allocation and documentation accuracy with text analysis
Overview of AI drug discovery progress in 2025, including the number of AI-discovered drug candidates, leading companies (Insilico Medicine, Recursion, Exscientia), and cost reduction data.
Company landscape diagram with pipeline count and cost comparison charts
Overview of virtual cell modeling in 2025: AI models simulating individual cell responses to drugs and interventions, enabling personalized medicine research and reducing animal testing requirements.
Virtual cell model capability diagram with application areas and text analysis
Data on AI applications in mental health in 2025: chatbot therapy tools deployment, efficacy data from clinical studies, crisis detection capabilities, and ethical concerns.
Efficacy study comparison chart with deployment statistics and ethical concern callout boxes
Analysis of AI applications in precision genomic medicine in 2025: polygenic risk score prediction, pharmacogenomics, rare disease diagnosis from genomic data, and AI-guided gene therapy design.
Capability comparison table with clinical application examples and text analysis
Overview of AI clinical decision support systems (CDSS) deployment in 2025: sepsis prediction, deterioration alerts, drug interaction checking, and their impact on clinical outcomes.
Outcome improvement bar chart by CDSS type with deployment data and text analysis
Data on AI-assisted surgery in 2025: robotic surgery AI guidance, real-time skill assessment, operative image analysis, and outcomes data for AI-assisted procedures.
Outcome comparison chart for AI-assisted vs traditional surgery with text analysis
Analysis of AI in emergency medicine in 2025: triage assistance, chest pain risk stratification, stroke detection on CT, and ED patient flow optimization.
Performance metric chart by application type with deployment data and text analysis
Analysis of AI's impact on health equity in 2025: potential to expand diagnostic access in underserved areas, but also risk of perpetuating or amplifying existing disparities if trained on biased data.
Equity impact dual-column analysis with geographic access expansion map and bias risk chart
Overview of AI applications in global health and low-resource settings in 2025: mobile AI diagnostics in LMICs, AI for neglected tropical disease, and telehealth AI deployment.
Geographic deployment map with application type breakdown and cost-effectiveness data
Analysis of AI medical device regulation in 2025 across major jurisdictions: FDA's predetermined change control pathway, EU MDR for AI, China NMPA AI device approvals, and international harmonization efforts.
Regulatory framework comparison table with approval timeline and text analysis
Analysis of AI integration with EHR systems in 2025: predictive analytics embedded in EHRs, NLP for unstructured notes, AI-driven care gap identification, and interoperability challenges.
EHR AI integration capability chart with adoption rate and text analysis
Data on AI-powered wearable health devices in 2025: continuous glucose monitoring, cardiac rhythm detection, respiratory monitoring, and population-level health monitoring programs.
Device capability comparison chart with clinical validation data and deployment statistics
Overview of AI patient communication tools in 2025: AI chatbots for patient education, medication adherence support, appointment scheduling, and post-discharge follow-up.
Tool type comparison chart with efficacy data and patient satisfaction survey results
Survey data on clinician attitudes toward AI in 2025: trust levels by specialty, concerns about liability, time to adopt new AI tools, and views on AI replacing versus augmenting clinical roles.
Survey result bar chart by specialty and concern category with text analysis
Overview of AI in medical education in 2025: AI clinical case simulators, AI-assisted board exam preparation, virtual patient interactions, and AI-guided surgical skills training.
Adoption rate chart with learning outcome data and tool type breakdown
Data on AI applications in hospital operations in 2025: patient flow optimization, bed management, staffing prediction, supply chain management, and revenue cycle AI.
Efficiency gain bar chart by operational domain with deployment statistics
Overview of AI in preventive medicine in 2025: population health risk stratification, cancer screening optimization, lifestyle intervention personalization, and public health AI tools.
Risk stratification performance chart with screening optimization data and text analysis
Analysis of how AI is transforming clinical trials in 2025: synthetic control arms, patient matching, endpoint prediction, and real-world evidence generation, with cost and timeline impact data.
Cost and timeline reduction chart with trial design methodology diagram and text analysis
Data on AI's impact on rare disease diagnosis in 2025: diagnostic time reduction, whole genome sequencing analysis AI, phenotype-to-genotype matching, and diagnosis odyssey shortening.
Diagnostic time comparison chart with case study examples and text analysis
Overview of AI in pharmacy and medication management in 2025: automated dispensing AI, drug interaction checking, personalized dosing optimization, and medication adherence tools.
Application capability chart with outcome data and text analysis
Analysis of AI's impact on healthcare costs in 2025: documented cost reductions in radiology, documentation, and operations, against implementation and integration costs.
Cost impact analysis chart with ROI estimates by application type
Performance data for AI dermatology systems in 2025: melanoma detection accuracy versus dermatologist, acne and skin condition classification, and AI-assisted teledermatology deployment.
Accuracy comparison chart by condition type with deployment statistics and text analysis
Analysis of AI in cardiology in 2025: ECG interpretation AI, echocardiography analysis, cardiac MRI segmentation, and AI-guided heart failure management.
Performance benchmark chart by cardiology task with clinical deployment data
Comprehensive overview of AI in oncology in 2025: tumor detection, treatment response prediction, radiation planning optimization, genomic biomarker identification, and precision cancer medicine.
Multi-application capability matrix with clinical evidence summary and text analysis
Data on AI applications in infectious disease in 2025: pathogen genomic surveillance, outbreak detection algorithms, antimicrobial resistance prediction, and pandemic preparedness AI tools.
Surveillance capability chart with deployment map and text analysis
Analysis of medical data infrastructure supporting AI in 2025: federated learning deployments, synthetic data for training, data sharing frameworks, and the role of national health data programs.
Infrastructure landscape diagram with federated learning deployment map and text analysis
Discussion of key ethical challenges in AI medicine in 2025: algorithmic accountability in diagnosis, patient consent for AI tools, liability when AI errs, and transparency requirements.
Ethical challenge taxonomy with regulatory response comparison and text analysis
Chapter 6 summary dashboard with key statistics: 258 FDA approvals, 300K+ clinicians using AI scribes, 85.5% MAI-DxO accuracy, AI drug candidates in Phase II/III trials.
Summary dashboard with large-number statistics and chapter blue-themed graphics
Reference notes and further reading for Chapter 6 with FDA data sources, clinical study citations, and curated reading list organized by medical specialty and AI application.
Two-column endnote format followed by structured reading list
Full-bleed chapter opener with forest green background, large white Chapter 7 numeral, title 'Education', and brief overview of AI's growing influence on learning, teaching, and academic institutions.
Full-bleed forest green background, oversized chapter number, white title and description
Introduction to Chapter 7 presenting AI in education as a pivotal issue in 2025, with headline findings: 80% of students using generative AI, 11% CS enrollment decline, 82% growth in AI master's programs.
Single-column introduction with headline statistics callout boxes and section overview
Line chart showing CS enrollment trends at top US universities from 2015-2025, including an 11% decline in 2024-2025, interpreted as a market saturation signal for traditional CS degrees.
Multi-institution enrollment line chart with year x-axis and enrollment count y-axis, 11% decline annotation
Bar chart showing growth in AI/ML-specific graduate programs from 2018-2025, with 82% growth in AI master's programs and new AI undergraduate concentrations at major universities.
Bar chart of program count by year with institution type breakdown and geographic distribution
Survey data showing 80% of college students report using generative AI for coursework in 2025, broken down by use type: writing assistance, research, coding, studying, and problem-solving.
Usage type bar chart with frequency distribution and demographic breakdown
Data on academic integrity concerns related to student AI use in 2025: detected AI-generated submissions, institutional policy evolution, and student views on AI use ethics.
Policy adoption chart with integrity incident data and student opinion survey
Survey data on faculty attitudes toward AI in 2025: 60% allow AI use with restrictions, varying policies by discipline, and faculty concerns about assessment validity.
Faculty survey bar chart by discipline with policy category breakdown and text analysis
Experimental evidence on AI tutoring system efficacy in 2025: Khan Academy Khanmigo, Duolingo Max, and similar AI tutors showing documented learning gains in controlled studies.
Learning gain comparison chart by tool with study methodology notes and text analysis
Data on AI tutoring tool deployment in 2025: number of students using AI tutoring platforms, geographic reach, subject area coverage, and usage frequency patterns.
Adoption chart with geographic map and subject area breakdown
Analysis of K-12 AI curriculum adoption worldwide in 2025, with 30+ countries implementing national AI/CS curriculum standards and data on teacher AI training availability.
World map of K-12 AI curriculum adoption with country count bar chart and text analysis
Content analysis of K-12 AI curriculum standards in 2025: what AI topics are taught (machine learning concepts, ethics, data literacy, prompt engineering) and at which grade levels.
Curriculum topic matrix by grade level with country comparison and text analysis
Data on online AI learning platform growth in 2025: Coursera and edX AI courses up 50%+, new AI certification programs from major tech companies, and completion rate analysis.
Growth trend chart with platform comparison and certification type breakdown
Analysis of how top universities are integrating AI research findings into teaching in 2025, including new AI ethics curricula, interdisciplinary AI programs, and AI research center growth.
University AI program comparison chart with research-teaching integration model diagram
Overview of AI literacy education initiatives in 2025: general public AI literacy programs, government-funded AI upskilling, workplace AI training, and the growing AI literacy certification market.
Program type overview with reach metrics and text analysis
Data on AI use in higher education administration in 2025: AI in student admissions, enrollment management, financial aid optimization, and student success prediction systems.
Application adoption chart with bias audit results and text analysis
Analysis of AI writing assistance use in academic contexts in 2025, including effects on writing quality, learning outcomes for writing skills, and evolving academic norms.
Study results comparison chart with writing quality assessment and text analysis
Overview of AI accessibility tools in education in 2025 benefiting students with disabilities: real-time captioning, AI reading assistance, personalized learning adaptations, and speech-to-text tools.
Accessibility tool capability chart with adoption data and student outcome evidence
Data on teacher and educator AI training programs in 2025: professional development participation rates, training content quality, and educator confidence in using AI tools.
Training adoption chart with confidence survey and content coverage comparison
Analysis of AI's potential impact on educational equity in 2025: personalized learning reducing achievement gaps, but device and connectivity access gaps creating new divides.
Equity impact dual analysis with access gap map and learning outcome data
Data on AI tools specifically supporting STEM education in 2025: AI math tutors, coding education platforms, science simulation tools, and lab safety AI applications.
Tool type comparison chart with learning outcome evidence and adoption data
Analysis of AI-powered language learning applications in 2025: Duolingo, Rosetta Stone AI features, and other platforms showing learning acceleration data and engagement metrics.
Learning rate comparison chart with engagement metrics and text analysis
Overview of national AI education policies in 2025: UNESCO's AI education guidance, EU AI education strategy, US AI literacy executive orders, and OECD AI education framework.
Policy framework comparison table with geographic adoption map and text analysis
Analysis of AI's impact on research university priorities in 2025: AI research center funding, industry-academia AI partnerships, AI patent activity from universities, and AI-related research revenue.
Research funding comparison chart with partnership data and text analysis
Chapter 7 summary dashboard presenting headline education statistics: student AI use rates, enrollment trends, AI tutoring efficacy, and K-12 curriculum adoption.
Summary dashboard with large-number statistics and chapter forest green-themed graphics
Reference notes and further reading for Chapter 7 including enrollment data sources, AI tutoring study citations, and curated reading list organized by education sector.
Two-column endnote format followed by structured reading list
Full-bleed chapter opener with dark purple/indigo background, large white Chapter 8 numeral, title 'Policy and Governance', and brief overview of AI regulatory and policy developments in 2025.
Full-bleed dark purple/indigo background, oversized chapter number, white title and description
Introduction to Chapter 8 framing 2025 as a watershed year for AI policy globally, with headline statistics: $20.4B US federal AI investment, 25 AI bills signed, 46 countries with national AI strategies.
Single-column introduction with headline statistics callout boxes and section overview
Comprehensive timeline visualization of major AI policy milestones from 2016 to 2025, from the Obama AI report (2016) through EU AI Act implementation (2025), organized by geography.
Horizontal multi-track timeline with country/region lanes, milestone markers, and brief annotations
World map showing the 46 countries that have adopted formal national AI strategies by 2025, with adoption date shading and regional clusters of AI governance activity.
World choropleth map with adoption date shading and country name callouts for major strategies
Analysis of US federal AI investment reaching $20.4 billion in 2025, broken down by agency (DOD, DARPA, NSF, NIH, DOE, NIST) and priority research areas.
Agency breakdown bar chart with priority area analysis and year-over-year growth annotation
Analysis of US AI-related legislation in 2025: 25 AI bills signed into law at federal level, topic breakdown (safety, procurement, workforce, deepfakes, national security), and state-level AI bills.
Legislative topic bar chart with timeline of signed bills and state-level activity map
Summary of major US executive branch AI actions in 2025: NIST AI Safety Institute activities, White House AI policy directives, and federal agency AI implementation mandates.
Policy action timeline with text analysis of key executive orders and agency responses
Analysis of US sector-specific AI regulations in 2025: FDA AI medical device guidance, FTC AI-in-advertising rules, CFPB AI credit scoring guidance, and EEOC AI hiring guidance.
Regulatory matrix by sector and agency with text analysis
Overview of the EU AI Act implementation milestones reached in 2025, including the prohibited practices ban that took effect February 2025, with a risk-tier classification diagram.
Risk tier pyramid diagram with implementation timeline and prohibited practices list
Detailed breakdown of EU AI Act risk categories: unacceptable risk (prohibited), high risk, limited risk, and minimal risk, with examples of AI applications in each category.
Four-category diagram with examples per risk level and regulatory obligation summary
Analysis of compliance requirements for high-risk AI systems under the EU AI Act in 2025: technical documentation, human oversight, data governance, and conformity assessment obligations.
Compliance checklist diagram with implementation cost estimates and sector-specific guidance
Data on EU AI Act compliance status among large enterprises in 2025, showing early adoption by large firms, compliance challenges for SMEs, and enforcement body establishment progress.
Compliance rate bar chart by firm size with enforcement progress metrics and text analysis
Analysis of the General Purpose AI Code of Practice under the EU AI Act, covering transparency requirements, systemic risk assessments, and how frontier AI developers are adapting.
Code of Practice obligations diagram with frontier model developer compliance status table
Overview of China's national AI policy in 2025: the Next Generation AI Development Plan implementation, $15B+ AI investment commitment, AI industry standards, and regulatory guidelines for generative AI.
Policy framework diagram with investment chart and regulatory timeline
Analysis of China's Generative AI Regulation (2023, updated 2025) provisions: content filtering requirements, watermarking mandates, real-name registration, and compliance enforcement.
Regulation provision summary with compliance burden analysis and text analysis
Overview of UK AI policy in 2025: AI Safety Institute evolution, the pro-innovation regulatory approach, AI sector-specific guidance, and post-Brexit AI governance divergence from EU.
Policy approach comparison (UK vs EU) with AI Safety Institute activity summary
Analysis of G7 Hiroshima AI Process outcomes and G20 AI governance framework adoption in 2025, including implementation status of the Hiroshima Code of Conduct for advanced AI developers.
Governance framework summary with implementation status indicators and country adoption table
Summary of the Paris AI Safety Summit (February 2025) outcomes: key commitments, participating nations, working group results, and the follow-up schedule for the global AI safety process.
Summit outcome summary with participating nation list and commitment tracking table
Analysis of UN AI governance activities in 2025: the Global Digital Compact, UN AI resolution adoption, the UN Secretary-General's AI Advisory Board, and the debate over a new UN AI agency.
UN governance activity timeline with text analysis and country position comparison
Analysis of the emerging AI sovereignty concept in 2025, defining its five dimensions: compute sovereignty, data sovereignty, talent sovereignty, standards sovereignty, and geopolitical AI independence.
Five-dimension framework diagram with country assessment per dimension and text analysis
Data on national efforts to build sovereign AI compute infrastructure in 2025: EU AI gigafactory plans, India's national AI compute grid, UAE's Falcon AI infrastructure, and others.
Compute investment comparison chart with geographic deployment map and text analysis
Analysis of AI-related export control policies in 2025: US semiconductor export restrictions expanded, chip-to-China rules, AI software export controls, and allied country coordination.
Export control timeline with country impact analysis and technology restriction map
Overview of AI standards development activity in 2025: ISO/IEC AI standards, IEEE AI ethics standards, NIST AI standards, and international standards harmonization efforts.
Standards landscape diagram with adoption tracking chart and text analysis
Comparative analysis of AI liability frameworks emerging in 2025: EU AI Liability Directive, US product liability approaches for AI, and industry self-regulatory liability frameworks.
Liability framework comparison table with jurisdiction-by-jurisdiction analysis
Analysis of AI's impact on democratic processes in 2025: AI in elections (deepfakes, voter targeting, AI-generated political content), regulatory responses, and international elections monitoring.
Impact analysis diagram with regulatory response timeline and text analysis
Cross-national comparison of public AI investment commitments in 2025: US $20.4B, EU €20B+ AI Act enforcement funding, China $15B+, UK £1.5B, Canada C$2.4B, and others.
Comparative investment bar chart by country with text analysis
Overview of AI applications in national security in 2025: intelligence analysis, cyber defense, logistics optimization, autonomous systems, and the US-China AI competition in defense.
Application domain overview with defense AI investment data and text analysis
Analysis of US state-level AI regulation in 2025: California AI bills, New York AI hiring rules, Colorado AI consumer protection, and the patchwork vs federal preemption debate.
State-by-state regulation activity map with bill count and topic breakdown
Analysis of government AI procurement policies in 2025: US federal AI acquisition rules, EU public sector AI requirements, and standards for responsible government AI procurement.
Procurement policy comparison table by jurisdiction with text analysis
Overview of AI-specific healthcare regulatory developments in 2025 beyond FDA device approvals: CMS AI coverage decisions, HIPAA-AI guidance, AI in clinical practice guidelines.
Regulatory development timeline with text analysis of key decisions
Analysis of AI regulation in financial services in 2025: SEC AI trading rules, banking regulator AI model risk guidance, CFPB AI in lending oversight, and global financial AI standards.
Regulatory framework comparison by financial sector with text analysis
Overview of AI labor regulation in 2025: EU AI Act worker surveillance provisions, US NLRB AI in workplace guidance, EEOC AI in hiring enforcement, and union AI negotiating demands.
Labor regulation comparison table with text analysis
Analysis of early evidence on AI policy effectiveness in 2025: which regulatory approaches are showing compliance results, unintended consequences of regulations, and policy gaps.
Policy effectiveness assessment matrix with text analysis
Overview of AI-specific privacy regulatory developments in 2025: GDPR AI guidance, US federal privacy bill AI provisions, data minimization for AI training, and biometric data restrictions.
Privacy regulatory comparison table with text analysis
Overview of civil society and NGO activity in AI policy in 2025: advocacy organizations, public interest AI research, coalition campaigns, and their influence on AI regulation.
Landscape diagram of civil society AI organizations with policy influence analysis
Chapter 8 summary presenting headline policy figures: US federal AI investment, global legislation counts, EU AI Act milestones, AI summit outcomes, and the AI sovereignty concept emergence.
Summary dashboard with large-number statistics and chapter purple/indigo-themed graphics
Concluding analysis of Chapter 8 identifying key governance trends: regulatory fragmentation risk, race to the top/bottom dynamics, AI sovereignty tensions, and the challenge of keeping pace with AI progress.
Text conclusion with key tension diagram and forward-looking analysis
Reference notes for Chapter 8 with policy document citations, legislative data sources, and investment figure methodology notes.
Two-column numbered endnote format
Curated further reading for Chapter 8 organized by policy region and topic: US policy, EU policy, China policy, international governance, and AI sovereignty.
Structured reading list with region and topic headers
Full-bleed chapter opener with blue-teal background, large white Chapter 9 numeral, title 'Public Opinion', and brief overview of global public attitudes toward AI surveyed in 2025.
Full-bleed blue-teal background, oversized chapter number, white title and description
Introduction to Chapter 9 framing the state of global public opinion toward AI in 2025: 59% global optimism from Ipsos, notable anxiety about disinformation, and significant expert-public attitude gaps.
Single-column introduction with headline statistics callout boxes and section overview
World map of global AI sentiment in 2025 showing the percentage expressing optimism about AI by country, with higher optimism in emerging economies and lower optimism in Europe.
World choropleth map of AI optimism with color gradient scale and country callout boxes
Detailed Ipsos survey results on AI attitudes in 2025 across 30+ countries: optimism, worry, trust, and awareness levels, broken down by country and world region.
Multi-country bar chart for each attitude dimension with regional aggregates
Regional deep-dive on AI sentiment: China and India showing highest optimism (80%+), Western Europe most cautious, US moderate, and Africa showing high optimism tied to development hopes.
Regional grouped bar chart with optimism/worry balance and text analysis
Comparison of AI attitude surveys for AI experts versus general public in 2025, showing experts more concerned about long-term risks while the public focuses on near-term economic impacts.
Side-by-side comparison chart with concern category breakdown for experts vs public
Survey data on public trust in various institutions to regulate AI fairly: EU institutions most trusted globally, tech companies least trusted, national governments mixed results.
Horizontal bar chart of trust levels by institution type across multiple countries
US-specific public opinion data on AI: only 31% of Americans trust the government to regulate AI well, 68% worried about AI-generated misinformation, and views on AI in various government roles.
US opinion bar chart with trust and concern breakdowns and text analysis
Analysis of AI anxiety sources in 2025: job loss concern, misinformation fear, privacy worry, autonomous weapons concern, and existential risk — broken down by country and demographic.
Stacked bar chart of anxiety sources by country group with demographic breakdown insets
Data on public AI awareness and understanding levels in 2025: rising awareness but limited technical understanding, with awareness-understanding gap analysis by education level and age.
Awareness vs understanding comparison chart with education and age breakdown
Survey data comparing AI attitudes across generations (Gen Z, Millennials, Gen X, Boomers) in 2025: younger generations more optimistic and higher daily AI usage, older generations more cautious.
Generational comparison bar chart for multiple attitude dimensions
Analysis of the correlation between education level and AI attitudes in 2025: higher education associated with more nuanced views rather than simple optimism or pessimism.
Correlation chart of education level versus AI attitude dimensions with text analysis
Analysis of the relationship between personal AI tool use and attitudes toward AI in 2025: regular users significantly more positive about AI's benefits, but also more aware of risks.
Use frequency vs attitude comparison chart with text analysis
Survey data on AI attitudes by income level in 2025: lower-income respondents more concerned about job displacement, higher-income respondents more optimistic about AI's economic benefits.
Income level attitude comparison chart with job displacement concern breakdown
Analysis of gender differences in AI attitudes in 2025: men slightly more optimistic on average, women more concerned about privacy and employment impacts, with variation by country.
Gender comparison bar chart across attitude dimensions with country variation insets
Line charts tracking AI attitude trends over time from 2019-2025: optimism rising through 2022, plateauing with more mixed views as AI becomes more personally relevant in 2024-2025.
Multi-line longitudinal chart with attitude category lines and event annotation
Global survey data on AI job displacement concerns in 2025, showing about half of global respondents worried about AI displacing their own job, varying significantly by sector and country.
Global concern bar chart by country with sector-specific variation analysis
Data on the extent to which people recognize AI in their daily products and services in 2025, showing high usage of AI tools but limited recognition that underlying technology is AI.
Usage vs recognition comparison chart with product category breakdown
Analysis of survey data on AI's perceived impact on social cohesion in 2025: concerns about AI exacerbating polarization, inequality, and reducing human connection.
Concern category bar chart with polarization and inequality callouts and text analysis
Survey data on public views about AI ethics in 2025: what values should guide AI development (safety, transparency, fairness, privacy) and who should have authority over AI ethics decisions.
Ranked values bar chart with authority preference breakdown and text analysis
Survey data on public concerns about AI's impact on democratic values in 2025: election integrity, free speech, surveillance, and the rule of law, with significant concern levels globally.
Democratic concern category bar chart with country comparison and text analysis
Chapter 9 summary presenting headline public opinion findings: global optimism rate, trust rankings, top AI anxieties, and key demographic divides in AI attitudes.
Summary dashboard with large-number statistics and chapter blue-teal themed graphics
Reference notes for Chapter 9 with survey data source citations and curated further reading on public AI attitudes organized by topic and methodology type.
Two-column endnote format followed by structured reading list
Overview of the AI Index 2026 research methodology, explaining data collection approaches across all nine chapters, the definition of 'AI' used in the report, and general methodological principles.
Structured methodology overview with section-by-section approach diagram
Detailed methodology for the R&D chapter: publication data sources (Semantic Scholar, Dimensions), patent data (WIPO, USPTO), model release tracking criteria, and compute data collection.
Methodology description sections with data source tables and limitation notes
Methodology notes for the Technical Performance chapter: benchmark selection criteria, data retrieval from Papers With Code and leaderboards, human baseline collection, and benchmark inclusion/exclusion rules.
Methodology description sections with benchmark selection criteria table
Methodology for the Responsible AI chapter: AIAAIC incident data collection, bias study inclusion criteria, safety research publication search methodology, and governance framework coding procedures.
Methodology description sections with inclusion/exclusion criteria
Methodology for the Economy chapter: investment data from NetBase Quid, job posting data from Lightcast, wage data sources, productivity study selection criteria, and firm adoption survey methodology.
Methodology description with data source reference table and limitation notes
Methodology for the Science and AI chapter: science publication AI-assisted classification methodology, domain-specific search criteria, and clinical trial data collection from ClinicalTrials.gov.
Methodology description with domain criteria and classification explanation
Methodology for the Medicine and AI chapter: FDA device data collection, clinical AI study inclusion criteria, diagnostic accuracy aggregation methods, and AI scribe adoption data sources.
Methodology description sections with FDA data approach and study criteria
Methodology for the Education chapter: enrollment data sources (IPEDS, international equivalents), student survey methodology, faculty survey design, and K-12 curriculum data collection approach.
Methodology description with survey design notes and data source references
Methodology for the Policy chapter: legislative database sources, AI bill identification criteria, investment figure compilation sources, national strategy coding methodology, and policy timeline construction.
Methodology description sections with legislative data source references
Methodology for the Public Opinion chapter: survey data sources (Ipsos, Edelman Trust Barometer, Stanford survey), survey sampling criteria, cross-national comparability notes, and attitudinal measure definitions.
Survey methodology description with sampling criteria and comparability notes
Glossary of key terms and data definitions used throughout the AI Index 2026, including definitions for 'foundation model', 'AI incident', 'training compute', 'benchmark saturation', and other core concepts.
Alphabetical glossary layout with term and definition entries
Comprehensive reference table of all AI benchmarks cited in the report, including benchmark name, domain, task type, human baseline, and year introduced.
Multi-column reference table sorted by chapter then benchmark name
Reference table of all notable AI models mentioned in the AI Index 2026, including model name, organization, release date, modality, and access type (open/closed).
Multi-column reference table sorted by organization and release date
Reference table of all AI policies, regulations, and legislation cited in the report, organized by country, including policy name, type, year, and status.
Policy reference table sorted by country and year
Supplementary investment data tables providing detailed private AI investment figures by country, sector, and year from 2013-2025 as referenced in Chapter 4.
Data table with country rows, year columns, and investment amount cells
Supplementary publication data tables providing detailed AI publication counts by country, domain, and year from 2010-2025 as referenced in Chapter 1.
Data table with country/domain rows and year columns
Supplementary tables presenting complete survey results from public opinion surveys cited in Chapter 9, including country-level breakdowns not shown in the main chapter.
Survey data tables with country rows and attitude metric columns
Supplementary tables of FDA AI-enabled medical device approval data cited in Chapter 6, including specialty, device type, 510(k)/PMA pathway, and applicant organization.
FDA device data table with multiple attribute columns
Bibliography entries A through D for all sources cited in the AI Index 2026, formatted in academic citation style with author, title, publication venue, and year.
Two-column bibliography format with alphabetical entries
Bibliography entries E through H for all sources cited in the AI Index 2026.
Two-column bibliography format with alphabetical entries
Bibliography entries I through L for all sources cited in the AI Index 2026.
Two-column bibliography format with alphabetical entries
Bibliography entries M through P for all sources cited in the AI Index 2026.
Two-column bibliography format with alphabetical entries
Bibliography entries Q through T for all sources cited in the AI Index 2026.
Two-column bibliography format with alphabetical entries
Bibliography entries U through Z completing the full works cited list for the AI Index 2026, followed by web resources and data platform references.
Two-column bibliography format with alphabetical entries followed by web resources section
Acknowledgment of the Stanford HAI research team members who contributed to the AI Index 2026, including research leads for each chapter and data analysts.
Structured acknowledgment list by chapter with contributor names and roles
Acknowledgment of external contributors, data providers, survey partners (Ipsos, NetBase Quid, Lightcast), and expert reviewers who contributed data or reviewed chapter drafts.
Acknowledgment list by contributor type with organization affiliations
Listing of the Stanford HAI AI Index Advisory Board members who provided guidance for the 2026 report, with names, titles, and institutional affiliations.
Advisory board member listing in alphabetical order with titles and affiliations
Extended case study summaries for 20 significant AI incidents in 2025 tracked in the AIAAIC database, providing detailed context beyond what appears in Chapter 3.
Case study cards in two-column grid with incident name, date, category, and summary
Detailed AI profiles for the US and Canada covering R&D output, investment, policy, talent, and adoption metrics with year-over-year comparisons.
Country profile cards with radar charts and key metric tables
Detailed AI profiles for the UK and major EU countries (Germany, France, Netherlands) covering R&D, investment, policy, talent, and adoption metrics.
Country profile cards with radar charts and key metric tables
Detailed AI profile for China covering R&D output, private investment, national AI policy, talent pool, model releases, and adoption metrics with year-over-year comparisons.
Country profile card with radar chart, key metric table, and policy highlight section
Detailed AI profiles for India and key Southeast Asian AI economies (Singapore, South Korea, Japan) covering R&D, investment, policy, and adoption metrics.
Country profile cards with radar charts and key metric tables
AI profiles for key Middle East AI actors (UAE, Saudi Arabia) and emerging African AI economies, with AI investment, national strategy, and talent development data.
Country profile cards with key metric summaries and national strategy highlights
Comprehensive tables of AI technical benchmark results for all models and benchmarks cited in Chapter 2, providing full data behind the chapter's charts and comparisons.
Large multi-column data tables with model rows and benchmark score columns
Supplementary enrollment, program, and survey data tables for Chapter 7, including institution-level CS enrollment data and full faculty/student survey cross-tabulations.
Data tables with institution and demographic rows
Detailed investment data tables providing company-level, sector-level, and geographic breakdown of private AI investment for 2023-2025 as referenced in Chapter 4.
Multi-level investment data tables by geography, sector, and year
Supplementary data tables for AI safety research publication analysis in Chapter 3, including search query protocols, keyword lists, and venue-level publication counts.
Safety research data tables with venue and year breakdown
Supplementary legislative data tables for Chapter 8, providing country-level counts of AI bills, legislative topic coding scheme, and key provisions summary for major AI laws.
Legislative data tables with country rows and bill category columns
Complete FDA AI-enabled medical device approval data through 2025 supplementing Chapter 6, with device-level detail on specialty, applicant, and approval pathway.
FDA device data tables with full detail columns
About page for Stanford Human-Centered AI Institute: mission, research focus areas, faculty and fellows, affiliated programs, and how to engage with HAI research and events.
Institutional overview with HAI program list and contact/engagement information
Back cover of the AI Index 2026 featuring Stanford HAI branding, report title, ISBN/DOI, website URL (aiindex.stanford.edu), and a brief one-paragraph summary of the report's purpose.
Back cover with report branding, summary paragraph, Stanford HAI logo, and URL
Common questions about this slide and the underlying presentation content.
The Stanford HAI AI Index Report 2026 is the 9th edition of the most comprehensive annual survey of artificial intelligence progress published by Stanford University's Human-Centered AI Institute. The 423-page report covers AI research and development, technical performance benchmarks, responsible AI, economic impact, science applications, medicine, education, policy and governance, and global public opinion — all grounded in original data analysis and hundreds of visualizations.
Key findings include: US private AI investment reaching \85.9 billion in 2025 (triple 2022 levels); AI agents now resolving over 50% of real software engineering issues on SWE-bench; the FDA approving 258 AI-enabled medical devices in 2025; 80% of college students using generative AI for coursework; 46 countries adopting national AI strategies; AI incidents tracked by AIAAIC increasing 1000% since 2019; and the EU AI Act's prohibited practices ban taking effect in February 2025.
The AI Index is published by Stanford University's Human-Centered AI (HAI) Institute, one of the world's leading AI research centers. Now in its 9th annual edition, the report is produced by a team of researchers with an advisory board of leading AI scientists, policymakers, and industry experts. It draws on primary data from dozens of established sources including Ipsos, NetBase Quid, Lightcast, the FDA, WIPO, and academic publication databases, making it a trusted reference for governments, corporations, and researchers worldwide.
The report contains 9 chapters: (1) Research and Development — model releases, publications, compute scaling; (2) Technical Performance — benchmarks across language, math, coding, science, multimodal; (3) Responsible AI — incidents, bias, safety, governance; (4) Economy — investment, labor, productivity, adoption; (5) Science and AI — biology, climate, materials, physics; (6) Medicine and AI — diagnostics, FDA devices, drug discovery; (7) Education — enrollment, student AI use, tutoring; (8) Policy and Governance — national strategies, EU AI Act, US policy; (9) Public Opinion — global attitudes, trust, anxiety. Front matter includes an executive summary with 10 key takeaways.
Yes, the Stanford HAI AI Index Report is made freely available to the public at aiindex.stanford.edu. The full 423-page PDF report, interactive data visualizations, and underlying datasets are provided at no cost, consistent with Stanford HAI's mission of ensuring AI's benefits and risks are widely understood.
The 9th edition is the most comprehensive yet, expanding coverage to include new sections on AI in science (Chapter 5), medicine (Chapter 6), and a dedicated chapter on public opinion (Chapter 9). The 2026 edition introduces new concepts like AI sovereignty (Chapter 8), adds coverage of the EU AI Act implementation, tracks the explosion in generative AI investment, and extends longitudinal data series back to 2010 in some cases — enabling a richer historical perspective on AI's progress trajectory than any prior edition.
The report contains hundreds of original data visualizations across 423 pages including line charts tracking AI capability and investment trends from 2010-2025, scatter plots comparing AI versus human performance on 30+ benchmarks, world maps showing global AI activity, stacked area charts of publication growth, capability comparison matrices for frontier models, policy timelines, survey data breakdowns by country and demographic, FDA approval trends, and many more. Each chapter includes endnotes and further reading, and the appendix provides full supplementary data tables.
Create Your Own Slides
Turn your ideas into professional presentations in seconds with 2slides AI.
Reference professional designs, choose your style, and generate slides with perfect text rendering. Powered by Nano Banana—start creating your presentation now.
Your AI Agent for slides. Save time, shine faster with intelligent presentation creation.
© 2026 2slides. All rights reserved.