ML Justice Lab leverages small, cost-effective language models to provide AI tools that perform on par with industry-leading solutions while being substantially more affordable and secure.

Our document analysis pipeline has been rigorously evaluated by legal professionals. Using Claude Haiku and open-source models like Mixtral, we've demonstrated performance that matches or exceeds solutions like CoCounsel while being dramatically more cost-efficient.

Importantly, our open-source models can be deployed with completely private endpoints, ensuring sensitive documents never leave your control. This combination of performance, affordability, and privacy makes powerful AI accessible to smaller non-profits and human rights organizations. The code, evaluations, and methodology are public and free. Click the links for more information.