Programmes · Open researchContributors welcome

Projects and tools

Open alignment research and tools, built with our fellows, researchers and partners.

Submit a proposalSee the projects
  • 1 openReady to pick up
  • 2 areasSafety and governance
  • PublicWrite-ups shared openly
Projects and tools Propose

Open projects

  • Technical AI SafetyOpen

    Testing Sarvam AI

    Testing Sarvam AI’s language models and benchmarking their alignment in Indian languages.

    Contribute to Testing Sarvam AI

Who it’s for

  • Fellows

    Turn your cohort project into open research.

    For graduates of our cohorts and fellowships.

  • Researchers

    Find people in India to build it with.

    For researchers with a question and room for help.

  • Engineers

    Build tools the safety community uses.

    For engineers who want their code to matter.

Research done in the open, with people who finish things.

An AI Safety India group together after a session in Bengaluru.
A speaker presenting at an AI Safety India event in Bengaluru.

How a project works

  1. 01Pick oneRead the brief and the open questions.
  2. 02Build it in the openWork with the project lead and other contributors.
  3. 03Write it upPublish what you found, so others can build on it.

Have a project or tool in mind?

Propose open alignment research or a tool.

Step 1 of 2Saved

Your idea

One clear question beats a long plan.

What is it?

What do you want to find out or fix, and why does it matter for safety?

A paper, a dataset, a working tool, a post.

What help do you need? (optional)

Good safety research is built together.

Submit a proposal
Fellows

Turn your cohort project into open research.

You finished a cohort or a fellowship and want to keep going. Pick up a project that matches what you studied, or bring the question you could not let go of.

Researchers

Find people in India to build it with.

You have a research question and more ideas than hands. Post it as a project, and work with contributors who read the brief and want to help.

Engineers

Build tools the safety community uses.

You write software and want it to count for AI safety. Projects need evaluation harnesses, datasets, benchmarks and the plumbing that makes research repeatable.

Programmes

Hubs