Open problems
A living list of concrete, tractable safety questions you can pick up — with data, baselines, and context to start fast.
Collaborative AI safety lab
Safe AI, built together.
Akxar Labs is an open, collaborative AI safety lab. We work in the open on alignment, evaluation, and human oversight — so the people building and using advanced AI stay meaningfully in control.
AlignmentKeeping capable systems pointed at human intent. EvaluationStress-testing reasoning, calibration, and failure modes. OversightHuman control, governance, and public benefit.
Featured research
Evaluating reasoning in frontier language models
A placeholder research note showing how Akxar Labs could report capability, reliability, and calibration measurements.
Read the researchCollaborate
Safety is a shared problem.
We treat safety research as a commons: open problems, open methods, and open tools. There are several ways to work with us — whether you are a researcher, an engineer, or a lab looking for partners.
Visiting researchers
Short residencies and remote collaborations for students and researchers who want to co-author work and share results.
Partner with us
Labs and teams can co-run evaluations, share infrastructure, and publish jointly under shared safety standards.