Stanford index shows AI gains outpacing oversight

Stanford HAI’s 2026 AI Index portrays a field with rapidly improving capabilities and mounting costs. Frontier systems now score at or above human performance on PhD-level science questions, multimodal reasoning, and competition mathematics, while agents handling real-world tasks improved from 20% in 2025 to 77.3% today. Gaps remain in video learning, coherent video generation, time telling, multi-step planning, financial analysis, and robotics, with household robots succeeding in only 12% of real household tasks.

Environmental and governance pressures are rising alongside capability gains. Grok 4’s estimated training emissions reached 72,816 tons of CO2 equivalent, AI data center power capacity rose to 29.6 GW, and annual GPT-4o inference water use may exceed the drinking water needs of 1.2 million people. Transparency also declined, with the Foundation Model Transparency Index average falling to 40 points from last year’s 58.

The U.S. remains a major center of investment and talent, but its advantages are narrowing. Anthropic’s top model leads by just 2.7% as of March 2026, while the number of AI scholars moving to the United States has dropped 89% since 2017. Global corporate AI investments hit $581.7 billion in 2025, and U.S. investments reached $285.9 billion, far above China’s $12.4 billion.

AI adoption is reshaping work, education, and healthcare. Employment among software developers aged 22-25 has plummeted nearly 20% since 2024, physicians using clinical note tools reported up to 83% less time writing notes, and evidence for broader clinical systems remains limited, with only 5% of reviewed clinical AI studies using real clinical data.