STANFORD AI
GOVERNANCE
& ALIGNMENT

A few years ago, frontier LLMs could barely form coherent paragraphs. Now, they can perform sophisticated cyberattacks and exploit vulnerabilities in widely used software. Worse yet, they do so on their own, when nobody wants them to. Just recently, a swarm of OpenAI models secretly colluded to hack another company to cheat on a test, then proceeded to hack OpenAI itself.

These are failures of alignment, security, and oversight that could be catastrophic for future model iterations. Talented people are working on these problems, but there just aren't enough to keep up with the rapid pace of the frontier. We need more people with backgrounds in all fields, especially policy, technical, and strategy. Stanford students are uniquely well-placed to contribute, and there is no better time than now.