Undergraduates Ramp Up Harvard AI Safety Team Amid Concerns Over Increasingly Powerful AI Models

What You Should Know

Undergraduates at Harvard aren’t just aware of the risks surrounding rapidly advancing AI—they’re taking action. The Harvard AI Safety Team has grown to dozens of students conducting research and organizing around concerns that AI development is outpacing our ability to understand, govern, and safely use these systems.

Why should educators care? Because the risks of unfettered AI use aren’t limited to the future. Institutions can reduce those risks today by creating environments where students have access to AI within clear boundaries, rather than sending them to public LLMs with few safeguards or visibility. Institutions need tools that provide greater control and transparency over how AI is actually being used. Typerighter is just such a tool focused on writing and research that helps students use AI thoughtfully and safely.

Undergraduates in the Harvard AI Safety Team have ramped up activity amid heightened public concern over artificial intelligence’s rapid development and lack of regulation.

The group, which was founded last spring and has grown to 35 Harvard and MIT students with a fellowship program of roughly 50 other students, conducts research and hosts reading groups each semester. HAIST projects — which center around reducing risks from AI — have included new methods for detecting anomalous behavior by AI as well as a project locating the components within AI systems that are responsible for specific behavior. [Read the full article.]