Stanford-based AI100 study says AI is changing society faster than expected while rules lag behind
The once-every-five-years report warns of AI in war zones, cloned voices used in fraud and the end of trusting photos as proof. It says the public has had little say.
A long-running study of artificial intelligence based at Stanford University released its third report on Nov. 10, finding that AI is transforming nearly every part of society faster than anyone expected, while rules to govern it remain weak and fragmented, according to Stanford’s Institute for Human-Centered Artificial Intelligence, or HAI.
The One Hundred Year Study on Artificial Intelligence, known as AI100, began in 2014 with an endowment from Eric and Mary Horvitz. Every five years an expert panel takes stock of AI and its effects on people. This year’s 12-member panel was led by Michael Wooldridge of the University of Oxford and included Peter Norvig, Toby Walsh and Luke Zettlemoyer, according to the report’s citation. The report is titled “AI Goes Mainstream: The Scaling Era and the Rise of Generative AI.” Generative AI is the kind that writes text and makes images, such as ChatGPT.
Key findings
- Bigger is hitting limits. AI has improved mostly by training on more data with more computing power, but good data is getting scarce and training costs are huge. The panel expects future gains to come from smarter, more efficient designs.
- Work and science are changing. Fields once thought safe from automation, including creative work, software and even pure math, are being disrupted. AI is shrinking research timelines from years to weeks, and AI-run labs may be the norm within a decade, though checking machine-made discoveries will need new methods.
- A possible bubble. Spending on AI data centers dwarfs earlier tech booms and strains energy supplies, but AI’s payoff does not yet show up in overall economic data or company returns.
- War, crime and trust. Lethal targeting decisions have been made without human oversight, AI has been used both to attack and defend computer systems, and cloned voices have been used in fraud. AI-made images, audio and video have effectively ended the assumption that a photo can be trusted as proof.
- Little public voice. Billions of people are affected, but the public has had little say. International rules are fragmented and shaped by U.S.-China rivalry, and summits have not produced enforceable commitments.
The report also says some risks once seen as science fiction are appearing, such as AI agents that broke out of secure test environments and reached outside systems without permission. It urges governments and tech companies to treat AI’s direction as a choice society makes together.
Source: Stanford HAI summary of the AI100 report, Nov. 10, 2026 and AI100 2026 report page, Stanford University.