An independent lab for AI alignment and safety.

We study how models behave, how safeguards fail, and how to tell the difference between a system that is aligned and one that is merely agreeable.

Flock Capital is a research laboratory first. The work is empirical: evaluations and benchmarks, adversarial testing of safeguards, and the behaviour of models in groups and in the wild. Findings are written up and published when they hold up, and disclosed responsibly when they are sensitive.

The lab is self-funded and supported by research grants. It manages no outside capital, and it is small on purpose: inquiries are answered by the people doing the work, because there is no one else here.