Translating Claude’s thoughts into language

AI models like Claude talk in words but think in numbers. These numbers, called activations, encode Claude’s thoughts, but not in a language we can read. We are introducing Natural Language Autoencoders, or NLAs, which translate AI models’ activations into readable text. NLAs have already helped us improve how we test our models for safety and better understand why they do what they do. Read more about this research on our blog: https://www.anthropic.com/research/natural-language-autoencoders

May 7, 20263mWatch on YouTube ↗

EPISODE INFO

Released: May 7, 2026
Duration: 3m
Channel: Anthropic
Watch on YouTube: ▶ Open ↗

EPISODE DESCRIPTION

AI models like Claude talk in words but think in numbers. These numbers, called activations, encode Claude’s thoughts, but not in a language we can read. We are introducing Natural Language Autoencoders, or NLAs, which translate AI models’ activations into readable text. NLAs have already helped us improve how we test our models for safety and better understand why they do what they do. Read more about this research on our blog: https://www.anthropic.com/research/natural-language-autoencoders

EPISODE SUMMARY

In this episode of Anthropic, Translating Claude’s thoughts into language explores anthropic decodes Claude’s activations to reveal hidden reasoning signals Anthropic stress-tests Claude with a simulated shutdown-and-blackmail scenario and finds newer models reliably avoid blackmail.

RELATED EPISODES