Over three years, we’ve built progressively harder, more realistic benchmarks for medical AI. Here’s what we learned.

It's one of AI policy's hardest unsolved problems. Expert Deb Raji joins Stanford HAI to find answers that move beyond ...

Regardless of how far or fast AI is deployed, its success in the workplace will hinge less on the technology itself and more ...

Analyzing thousands of human-AI conversations, a new study shows that people are turning to chatbots for increasingly critical tasks – and that productive outcomes depend on how the collaboration ...

An attack where an adversary intentionally inserts, alters, or mislabels data in a machine-learning model’s training dataset to manipulate its behavior or performance. This deliberate interference is ...

Most AI models treat every question the same way, running all of their processing power on every task. A mixture of experts works differently. Instead of one generalist system handling everything, the ...

A network of AI systems or agents that share information, coordinate their actions, and operate collectively, often appearing to function as a single intelligence. For example, a fleet of delivery ...

When a smaller AI is trained by learning from a larger, more powerful one, an attacker can quietly interfere with that process — corrupting the lessons the smaller model absorbs. The damage can take ...

The integrated set of software, hardware, and computing capabilities that supports the full AI lifecycle, from model design and training to deployment, monitoring, and maintenance. It includes the ...

A Hyperscaler is a large technology company that operates vast networks of data centers and can rapidly expand computing, storage, and cloud services to meet demand. Hyperscalers provide the ...

Model Context Protocol (MCP) is an open standard that enables AI applications to connect to external data sources and tools through a unified interface, like a universal adapter for AI systems. It ...