Explore how CTGT’s research is deepening our understanding of frontier intelligence.

Sep 17, 2026

Defensive Interpretability for Frontier Models

Where interpretability has to live when the adversary holds the same weights, and what it costs. The detection instrument is the cheapest object in the stack by six orders of magnitude.
READ MORE
More Blogs

Research, publications & industry insights

Sep 10, 2026
Research

Privileged Information in Self-Distillation: An Empirical Study

Aug 24, 2026
Research

Behaviorally Fingerprinting Ox Alpha's Provenance and Censorship

Aug 5, 2026
Research

DeepSeek’s Official V4 Flash Censors More Than Its Preview, Selectively

Jul 29, 2026
Research

What a Distilled Model Inherits From Its Teacher

Dec 9, 2025
Research

Introducing Mentat: Reliable Runtime Control for LLMs

Mar 14, 2025
Research

A Feature-Level Approach to Mitigating Bias and Censorship in DeepSeek-R1

Feb 20, 2025
Announcement

CTGT Raises $7.6M To Deepen Understanding of Frontier AI Behavior

Interested in partnering with CTGT?

contact us
Industries
Finance
Company
About us
Research
Contact us
Careers
Trust
Legal
Terms & conditions
Privacy policy
We use cookies
We use cookies to ensure our site functions properly, understand how it’s used, and improve your experience. Some cookies are essential, while others help us analyze usage and performance.

You can manage your preferences at any time. By clicking Accept, you agree to our use of cookies as described in our Privacy Policy.
Accept All
Reject all