AI Index 2025: 10 Charts Reveal Shifts in Power, Progress, and Risk

The 2025 AI Index Report, released by Stanford University’s Institute for Human-Centered Artificial Intelligence, paints a picture of a rapidly evolving landscape: smaller and more efficient models, rising incidents of AI misuse, and widening geopolitical divides in AI power and optimism.

Produced by an interdisciplinary steering committee, the annual report tracks progress in model performance, cost trends, investment, regulation, education, and real-world adoption.

Here are 10 key takeaways from this year’s findings:

Smaller Models, Bigger Impact: AI models are becoming dramatically more efficient. In 2022, a model needed 540 billion parameters to cross a 60% score on the MMLU benchmark. By 2024, Microsoft’s Phi-3-mini did it with just 3.8 billion—a 142-fold reduction in scale in just two years.

Costs Drop by Orders of Magnitude: The price of querying a GPT-3.5-level model fell from $20 per million tokens in 2022 to $0.07 by late 2024—a 280-fold reduction in just 18 months. Depending on the task, LLM inference costs are falling as much as 900 times per year.

China Closes the Quality Gap: While U.S. institutions still lead in AI model output, China is catching up fast in performance. Major benchmark scores between the U.S. and China narrowed from double-digit performance gaps in 2023 to near parity in 2024. China also leads globally in AI patents and publications.

AI Harms Are on the Rise: The AI Incidents Database recorded 233 harmful AI-related incidents in 2024—a 56% jump from the previous year. Cases included deepfake abuse and chatbot-related mental health crises.

AI Agents Gain Ground: New benchmarks show that AI agents now outperform humans by a factor of four in short tasks lasting up to two hours, but still fall behind in longer-duration challenges. In 32-hour challenges, humans still beat AI by 2 to 1. Nonetheless, AI agents already match human-level performance in specific tasks like code generation.

U.S. Leads in AI Investment: Private AI investment in the U.S. reached $109 billion in 2024—12x China’s $9.3 billion and 24x the UK’s $4.5 billion. In generative AI alone, the U.S. outpaced the EU and UK combined by $25.5 billion, up from a $21.1 billion gap in 2023.

AI Becomes Business Standard: Corporate AI adoption surged. In 2024, 78% of organizations reported using AI—up from 55% in 2023. Use of generative AI in business functions more than doubled, from 33% to 71%.

Health AI Booms at the FDA: The FDA has now approved 223 AI-enabled medical devices, up from just six in 2015. The sharp rise reflects growing momentum for AI in diagnostics, imaging, and medical decision support.

States Step In on Regulation: With limited federal action, U.S. states have taken the lead. 131 state-level AI laws were passed in 2024—more than double the previous year’s total. By contrast, few federal bills have made it into law.

Asia Shows Greater AI Optimism: Public attitudes vary globally. In China (83%), Indonesia (80%), and Thailand (77%), people see AI as more beneficial than harmful. In contrast, optimism is far lower in the U.S. (39%), Canada (40%), and the Netherlands (36%).

What This Means

The 2025 AI Index reveals a field both accelerating and fragmenting. AI is getting faster, cheaper, and more capable—but it's also growing harder to govern, especially with federal regulation lagging in the U.S. and nations like China rapidly advancing their own ecosystems.

The surge in AI-related incidents highlights the urgent need for safety frameworks, while state-level legislative action signals a shift toward decentralized policymaking. Meanwhile, falling costs and smaller model sizes could democratize AI access, but also raise the risk of misuse by bad actors.

On the global stage, the U.S. retains leadership in investment and innovation, but China’s rise in performance and publication shows a tightening race. And as businesses adopt AI faster than regulators can respond, questions around ethics, bias, and transparency will only become more urgent.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.