Skip to content

What We Do

We operate as a hub for the AI safety ecosystem. Our work is divided into three distinct pillars designed to shepherd the future of technology:

01

Innovative Research

We look where others aren't looking. While big labs focus on making models talk better, we focus on what goes on "under the hood."

  • Technical Safety: We study how to make AI robust against errors and manipulation.
  • Social Impact: We actively test models for harmful biases. For example, our work includes Project Trixie, an initiative dedicated to detecting and removing antisemitism and hate speech from AI outputs.
02

Empowering Talent(Grant Funding)

We believe the solution to AI safety won't come from just one laboratory. It requires a global effort. We provide financial grants and fiscal sponsorship to:

  • Independent researchers with breakthrough ideas.
  • Educational programs building the next generation of safety experts.
  • Engineers developing tools to visualize how AI "thinks."
03

Field Building

We are connecting the dots between policy advisors, technical experts, and the public. By fostering a community around safety, we ensure that resources go to the high-leverage projects that need them most.

What is Alignment?

AI alignment is the challenge of making sure AI systems actually do what humans want them to do, and don't pursue goals that conflict with our values or interests.

As AI systems become more capable, this challenge becomes more urgent. An AI system that is highly capable but not aligned with human values could cause significant harm, not out of malice, but simply because it pursues goals that don't account for what we actually care about.

Neglected Approaches

We fund what we call "neglected approaches": ideas that might actually work, but that almost nobody is seriously working on yet. This is usually because of profit incentives, which is why we're structured as a non-profit.

We believe that the greatest breakthroughs will come from exploring weird, overlooked, technically rigorous ideas that could establish alignment as a self-reinforcing property of advanced AI systems.

Protect the Future

AI is the ultimate lever of power. If we do not align it with human freedom, it will be used to manage human behavior. The nations that master this technology secure the right to remain free.

We work to ensure this power (and economic advantage) belongs to those who value the individual over the collective.

How can I fund alongside you?

We encourage fellow philanthropists to consider co-funding with us. No single organization will solve AI alignment. The problem is too vast and the stakes are too high to rely on any one approach.