A number in this year’s Stanford HAI AI Index Report is so interesting that it’s hard to put down once you read it. Six percent of teachers say that the AI policy at their school is not clear. Not not good enough or not enforced well enough; just really clear. This is the year when four out of five high school and college students in the U.S. are already using AI for schoolwork. The difference between what students are doing and what institutions are ready for feels less like a lag and more like a slow-moving structural failure.
Stanford’s Institute for Human-Centered Artificial Intelligence puts out the 2026 AI Index, which is now in its ninth edition. It covers a wide range of topics, such as model performance and compute infrastructure, as well as economic impact, education, policy, and public opinion. Most people agree that it is the most accurate annual picture of where AI really is, not where the news stories say it is. The numbers in this year’s edition carry more weight than usual because they show a field that has given up waiting for society to catch up.
Within three years, 53% of people had used generative AI, which is faster than the PC or the internet. However, the rate varies by country and is strongly linked to GDP per capita. Technology quickly became a part of everyday life, faster than institutions like schools, hospitals, and governments could keep up. The report makes it pretty clear how much that difference costs.
The progress is really impressive when it comes to skills. In 2025, industry made more than 90% of the most important frontier models. Some of these models now meet or beat human baselines on PhD-level science questions, multimodal reasoning, and competition math. A key coding test called SWE-bench Verified showed that performance went from 60% to almost 100% in just one year. There was a math competition, and Google’s Gemini Deep Think won first place. That’s not a figure of speech. These systems are competing, and winning, in areas that used to feel very human.
However, the report makes sure that the main numbers don’t say everything. The best model on ClockBench could read analog clocks correctly only 50.1% of the time, while humans can do it 90.1% of the time. AI agents got a lot better at structured computer tasks, but robots still fail about nine out of ten real-life household tasks. It looks like their intelligence is all over the place—they are very smart at some things and surprisingly weak at others. This inequality might be more important than it seems because it makes AI really hard to judge, especially for the teachers and administrators who have to make policy decisions about it.
Now we’ll talk about something that should worry the people who run schools. There are three things in this report that need more attention than they are getting. Four out of five high school and college students in the U.S. use AI for schoolwork right now, but only half of middle and high schools have policies in place for AI, and only 6% of teachers think those policies are clear. Students already know how to use this tech. Most of the time, the institutions that are supposed to help them learn don’t have a map.
Even though there is a lot of interest and money being put into AI educational tools, the 2026 AI Index finds that there is still not a lot of solid evidence on how AI-enhanced education affects learning. Some measured outcomes get better with AI tutoring tools in some situations, but not in others. In some cases, they make learning worse by reducing the productive struggle that builds understanding. Schools are being sold peace of mind. So far, the data doesn’t back it up.
Then there’s the workforce signal, which changes what teachers are really teaching their students. It’s now real that AI will change the way people work, and it’s mostly affecting young workers. The job market is already changing under the feet of the students sitting in classrooms right now. This report brings up the question of whether schools are preparing them for this change or even recognizing it, but it doesn’t fully answer it.
Starting with the 2025–26 school year, both China and the United Arab Emirates made AI education required. This marks the start of formal AI education at the national level. The United States does not have a similar national mandate. Even though the value of requiring a certain curriculum from the top down is questionable, it’s impossible not to notice the difference.
The report paints a bigger picture of amazing new technologies coming out faster than anyone can control them. These technologies affect the environment, politics, and the economy. Since 2017, 89% fewer AI scholars have moved to the US. In the last year alone, that number has dropped by 80%. There is a problem with the talent pipeline, and it will have long-term effects that go beyond any single benchmark score.
After reading all 400+ pages, it seems like the report is trying to say two things at once: it’s genuinely positive about what these systems can do, and it’s genuinely worried about how quickly they’re being added to systems that weren’t made to handle them. In particular, the 2026 AI Index is not a reason for educators to freak out. It’s a call to get caught up, and not just with promises. You need to have proof as well.
