About Us
Who We Are
Formation Research Ltd is a UK-based not-for-profit company limited by guarantee with 501(c)(3) equivalency determination through NGO Source. We are on a mission to research fundamental lock-in dynamics and implement high-impact interventions.
A lock-in is a situation where some feature of the world, typically a negative element of human culture, becomes stable for a long time. Formation Research focuses on interventions for particularly undesirable lock-in scenarios such as AI-enabled totalitarianism and extreme power concentration.
Current Research Focus
Secret Loyalties
A promising technical intervention direction for lock-in risk is secret loyalties research. A secret loyalty is an objective encoded in a language model that favours a specific actor or the goals of that actor, activated by a private activation condition.
Secret loyalties are a mechanism for AI-enabled authoritarianism, power concentration, and similar lock-in risks from AI systems. There is concrete technical research that can be done on understanding and mitigating secret loyalties with current systems.
In collaboration with researchers at
Lock-In Scenarios We Study
We focus on a range of undesirable lock-in scenarios, including:
AI-Enabled Totalitarianism
A human dictator may leverage AI systems to improve their own lifespan and implement mass worldwide surveillance, leading to a long-term stable totalitarian regime.
Extreme Power Concentration
Individuals at the forefront of the AI revolution may end up with high leverage over the technological and political trajectory of humanity, creating persistent monopolies over resources and labour.
Our Vision
Minimising Lock-In Risks
Reducing the likelihood that harmful, oppressive, or persistent elements of culture become stable, whether through human action or AI systems.
Promoting a Dynamic Future
- Continued technological and cultural evolution
- Economic growth
- Sustainable competition
- Improved individual freedom
Our Approach
Our research is defined by first-principles, bottom-up, collaborative, scientific and technical investigation into AI systems and their potential uses.
First Principles
Building conceptual models from fundamental understanding of physics and computation, testing assumptions before employing them.
Bottom-Up Research
Creating inside-view theoretical models based on simple facts about AI systems and game theory.
Collaborative
Working with AI safety organisations and conducting interdisciplinary research with think tanks and economists.
Scientific Method
Using conjecture, criticism through peer review, and error-correction to create fundamental knowledge about lock-in risks.
Technical Research
Creating applicable knowledge for real-world interventions and developing practical mitigations for lock-in risks.
Validated Learning
Continuously updating our research agenda and interventions based on evidence and reason.
Our Team
Alfie Lamerton
Founder
Alfie Lamerton founded Formation Research in 2025. He holds a computer science BSc and artificial intelligence MSc, and has worked as a software engineer, research assistant, and independent researcher. He has participated in several AI safety projects and been awarded several grants for his research.
