Stanford’s AI Index 2026 shows $285.9B U.S. lead and a safety gap

Stanford HAI’s 2026 AI Index, released on April 13, reads like a scorecard for an industry that keeps sprinting while its referees fall behind. The report says AI is moving deeper into the economy through better technical performance, faster adoption, and more capital, but the systems that measure safety, transparency, and governance are not keeping up. That framing matters because it shifts the conversation away from “which model won this week” and toward whether anyone can still independently describe what progress is costing and where risk is accumulating. The full report is worth reading directly in Stanford HAI’s 2026 AI Index.

The money numbers alone explain why the gap is widening. Stanford says U.S. private AI investment reached $285.9 billion in 2025, more than 23 times China’s $12.4 billion. The U.S. also produced 1,953 newly funded AI companies in 2025, more than 10 times the next closest country. That is the kind of lead that can accelerate infrastructure buildout, talent bidding, and product rollouts all at once. But the report also notes a softer undercurrent: the number of AI researchers and developers moving to the U.S. has fallen 89% since 2017, including an 80% drop in the last year alone.

The most uncomfortable section is the one on responsible AI. Stanford says almost every leading frontier model developer reports capability benchmarks, but reporting on responsible AI benchmarks remains spotty. At the same time, documented AI incidents climbed to 362 in 2025, up from 233 in 2024. The report adds a harder problem than simple under-reporting: recent research suggests improving one responsible-AI dimension, such as safety, can degrade another, such as accuracy. In other words, the measurement system is thin just as tradeoffs are becoming more complex.

That combination makes the 2026 Index feel less like a vanity compilation and more like a warning label for the next phase of deployment. Capital, open-source participation, and national AI strategies are all broadening who gets to build. But if independent measurement stays weaker than model marketing, the market will keep pricing speed more accurately than risk. That is a durable shift to watch in 2026, because governance gaps usually stay invisible right up until they start moving capital, procurement, and regulation.