Stanford’s One Hundred Year Study on Artificial Intelligence (AI100) has released its third five-year assessment, “AI Goes Mainstream: The Scaling Era and the Rise of Generative AI,” examining how AI has shifted from specialized research to an everyday technology used by billions.

The report says AI’s progress and influence have accelerated beyond expert expectations, especially after the 2023 “ChatGPT moment,” which the 2021 panel had not anticipated; the next five years could bring similarly surprising changes.

Scaling models with more data and computing power drove major advances, but rising costs and limited high-quality data are pushing research toward more efficient systems, stronger reasoning and additional computation when models produce answers.

The panel expects progress to rely increasingly on smarter algorithms, stronger context and memory, specialized reasoning, and tools or agents for complex tasks, while warning that AI capabilities remain uneven.

AI is reshaping creative work, programming and mathematics, while accelerating some scientific research from years to months or weeks; the changes also raise concerns about research quality, academic integrity and how AI may affect collective thinking and moral behavior.

Rahwan sees significant potential for AI to speed scientific discovery and support high-quality reviews of research, even as AI-generated papers put pressure on peer review.

Despite conspicuous investment and market valuations, the panel says AI’s effects are not yet clear in aggregate economic data or companies’ returns on investment.

Meanwhile, the build-out of AI data centers and chips is straining energy and freshwater supplies and increasing carbon emissions; Nvidia’s valuation exceeded $5 trillion in 2025, but the report questions whether massive infrastructure investments will pay off as public opposition to data-center construction grows.

The report warns that AI is creating cybersecurity risks, enabling both cyber defense and attacks, and has been used in active conflicts, including in lethal targeting decisions made without humans in the loop.

Synthetic images, cloned voices and video are weakening trust in recordings as evidence in journalism, law and democratic institutions, while the public has had little meaningful input into AI’s development and deployment.

Rahwan argues that AI is more than a tool: when people act on its recommendations, it can influence decisions and behavior, giving it a form of collective agency within society.

He says AI should be studied as an actor in social systems, with behavioral and social-science expertise needed to understand its effects on work, education, hiring and relationships.