Stanford’s 2026 index depicts Artificial Intelligence as a technology advancing at extraordinary speed while the systems around it lag behind. Top models continue to improve despite repeated predictions of a plateau, and adoption is accelerating across society faster than earlier waves such as the personal computer and the internet. At the same time, companies are racing to build infrastructure, benchmarks are losing usefulness, and policymakers are struggling to govern systems that remain poorly understood.
The competitive landscape is tightening between the US and China. Arena rankings show the two countries nearly tied in model performance, with leadership shifting among Anthropic, xAI, Google, OpenAI, and Chinese firms including DeepSeek and Alibaba. The US retains advantages in capital and infrastructure, with an estimated 5,427 data centers, while China leads in research publications, patents, and robotics. Transparency is also declining as leading companies increasingly withhold training code, parameter counts, and data-set sizes, making outside safety research more difficult.
Performance gains remain striking. SWE-bench Verified, a software engineering benchmark for Artificial Intelligence models, saw top scores jump from around 60% in 2024 to almost 100% in 2025. By some measures, leading systems now match or exceed human experts in PhD-level science, math, and language tasks, and in 2025, an Artificial Intelligence system produced a weather forecast on its own. Yet those gains are uneven. Robots succeed in only 12% of household tasks, and the report describes model behavior as jagged rather than consistently capable across domains.
The report also warns that the tools used to evaluate progress are breaking down. Some benchmarks contain major flaws, including a widely used math benchmark with a 42% error rate. Others can be gamed when models are trained on benchmark data, and strong scores often fail to predict real-world usefulness, especially for agents and robots. Independent testing is becoming more important as companies disclose less, particularly on responsible-Artificial Intelligence measures.
Economic and social effects are becoming more visible. Artificial Intelligence is now used by more than half of people around the world, an estimated 88% of organizations now use Artificial Intelligence, and four in five university students use it. A 2025 Stanford study found employment for software developers aged 22 to 25 has fallen nearly 20% since 2022. Research cited by the index says Artificial Intelligence is boosting productivity by 14% in customer service and 26% in software development. Public attitudes remain mixed: 59% of people think that it will provide more benefits than drawbacks, while 52% say that it makes them nervous. Governments are responding unevenly, from the EU AI Act’s first prohibitions to a record 150 Artificial Intelligence-related bills passed by US states, underscoring how regulation continues to trail the technology.
