Our research teams investigate the efficiency, security, and transparency of AI models, studying how artificial intelligence can be designed to operate safely and reliably at scale.
Foundation Models
Exploring methods to train and adapt AI models with improved efficiency and robustness.
Dedicated to uncovering how AI systems work internally. This team studies the representations, pathways, and decision processes of models, building the foundation for interpretability and trust.
Alignment Team
Focused on keeping AI behavior consistent with human intentions and values. The team develops methods to ensure models remain reliable, safe, and beneficial as capabilities grow.
Impact & Foresight
Studies the broader effects of AI on society, industry, and policy. This team works to anticipate challenges, guide ethical deployment, and ensure long-term positive outcomes.
Security & Resilience
Investigates the robustness of AI against failures, adversarial attacks, and misuse. Their mission is to design systems that stay dependable in real-world environments.
Research Teams
Transparency
We believe AI should not be a black box. Our research is dedicated to uncovering how models learn, reason, and generate outputs, so their decisions can be explained and trusted by those who rely on them.
Safety
AI must remain dependable under real-world conditions. We focus on building systems that resist failures, attacks, and misuse, ensuring stability and protection in critical environments.
Alignment
Intelligent systems should reflect human goals and values. We develop methods that guide AI behavior to stay useful, honest, and responsible as capabilities scale.
Impact Awareness
AI does not exist in isolation. Our research examines its wider effects on people, industries, and institutions, aiming to anticipate risks while steering progress toward lasting positive impact.