Projects

Jailbreaking Legal AI

Stress testing LLMs to find where they break down.

Judicial Education Think Tank

Brainstorming how judges can safely use AI.

In progress

Model Security & Guardrail Testing

Locally installed open models, sandboxed from any commercial platform, are used to test confidentiality across accounts, resistance to jailbreaking, and hallucination under legal-narrative pressure. Students with research-computing access assist.

Read about the research →