The State of AI Code Control 2026
Coding is solved. Its control is not.

We are an independent group of data architects and scientists studying how organizations control AI-generated code. Our first study, The State of AI Code Control 2026, analyzed 112 million commits from 340,000 repositories in 159 countries.
112M
commits analyzed
340,000
repositories, 159 countries
69%
use AI agents in their code work
75%
of agent pull requests merged without human review
What we do
Research
how AI-generated code is governed, published openly under CC BY 4.0.
Monitoring
exploring a platform that lets organizations follow control of their own code (private beta).
Talent
exploring the same method for validating developers' code history in recruitment.
Join the AgenticReview Argonauts
We are looking for people who want to shape how AI-generated code is governed - as sparring partners in general or in a specific research or development project.
Apply to joinWho we are
The group is led by Rufus Lidman (Fil. Lic.), data and AI architect with more than 20 years of leadership experience across four continents, together with data scientists, data architects and software engineers. The study was financed by seed funding from the founders.
Press
Media contact: Rufus Lidman, rufus.lidman@agenticreview.ai, +46 767-098 004. High-resolution figures may be republished under CC BY 4.0.
Figures for media