Open research in AI, for all of humanity.

Forall Labs is an open research community. Anyone can join, every result is published, and the work is chosen by who it helps.

A lattice of points, the lit ones forming the Forall Labs mark: the symbol for all, followed by an L

Research

We focus on AI itself, and apply it where it does the most good.

Four themes, each a place where openness is an advantage rather than a cost.

01

Open models and tools

Models, datasets and training recipes anyone can reproduce on modest hardware.

02

AI for people

Health, education, climate and public services, built with the communities who use them.

03

Safe and accountable AI

Evaluation, transparency and safety work that keeps powerful systems answerable to the public.

04

Embodied and efficient AI

Robots and agents that work in the open world, and training methods small labs can afford.

Read the research programme

What we value

The principles we hold ourselves to.

  1. 01

    Publish everything.

    Papers, code, data and models are released openly, so anyone can check them, reuse them and build on them. A closed result needs a reason.

  2. 02

    Choose work by who it helps.

    Real-world benefit and safety come before novelty, and before results that only look impressive.

  3. 03

    Decide together, in the open.

    No university, company or investor sets the agenda. Members propose the work and choose it together.

  4. 04

    Credit everyone who contributes.

    Every contributor is named. A title, an affiliation or a PhD is not required to do serious research here.

  5. 05

    Report what did not work.

    Negative results are results. Writing them up saves the next person the same months.

How it works

A project here starts with a proposal, not a permission.

Anyone proposes

Write up an idea: what it is, who it helps, what it needs. No affiliation required, and no committee to get past first.

A team forms

Proposals are shared openly so collaborators can find them. Projects that attract people get going.

Everything is published

Code, data and the paper go out openly, with every contributor named. Negative results get written up too.

Want to help build AI research that belongs to everyone?

Membership is free and open to anyone, anywhere.