studies how large language models work internally as a foundation for AI safety. They seek to understand what happens inside neural networks so that we can better trust the models we build.
- Team blog
- Visit
Safety through understanding
It’s hard to reason about the safety of neural networks without understanding what happens inside them. Our goal is to be able to explain large language models’ behavior, and to use that understanding to identify problems that are hard to see from the outside, such as situations where a model’s stated reasoning differs from what it’s actually doing.
An evolving approach
Our early work focused on mapping the internal mechanisms of AI models. Today, we increasingly apply our methods to real models to understand how they think and why they behave the way they do. We’ve developed techniques that allow us to understand models’ inner workings in plain language, making it easier to detect unsafe behavior or dishonest reasoning and safeguard against it.
Natural Language Autoencoders: Turning Claude’s thoughts into text
AI models like Claude talk in words but think in numbers. In this study, we train Claude to translate its thoughts into human-readable text.
Read moreA global workspace in language models
New interpretability research reveals an emergent mental workspace in Claude that holds internal thoughts that don’t appear in the model’s output.
Read moreEmotion concepts and their function in a large language model
All modern language models sometimes act like they have emotions. What’s behind these behaviors? Our interpretability team investigates.
Read more01
/
03
35
ARTICLES
Explore work from other teams
Alignment
47 posts
Works to understand the risks of AI models and develop ways to ensure future ones remain helpful, honest, and harmless.
Economics
25 posts
Studies and monitors the impact of AI systems on individuals, firms, and economies.
Engineering
27 posts
Builds the infrastructure and systems that make Claude reliable and trustworthy.
Frontier Red Team
33 posts
Stress-tests AI systems to understand the full extent of their current capabilities and anticipate what comes next.
Societal Impacts
24 posts
Studies how AI affects people, both individually and collectively.
Science
18 posts
Explores how AI can increase the pace of scientific progress through both research and collaborations.
Interpretability
35 posts
Studies how large language models work internally as a foundation for AI safety.
Alignment
47 posts
Works to understand the risks of AI models and develop ways to ensure future ones remain helpful, honest, and harmless.
Economics
25 posts
Studies and monitors the impact of AI systems on individuals, firms, and economies.
Engineering
27 posts
Builds the infrastructure and systems that make Claude reliable and trustworthy.
Frontier Red Team
33 posts
Stress-tests AI systems to understand the full extent of their current capabilities and anticipate what comes next.
Societal Impacts
24 posts
Studies how AI affects people, both individually and collectively.
Science
18 posts
Explores how AI can increase the pace of scientific progress through both research and collaborations.
Interpretability
35 posts
Studies how large language models work internally as a foundation for AI safety.