SAN FRANCISCO — AI research company OpenMind unveiled LexAI-4 on Wednesday, a large language model that scored in the 94th percentile on a full administration of the Uniform Bar Examination — a result that would qualify for bar admission in all 50 US states.
The achievement marks a significant leap from previous AI legal benchmarks and has renewed urgent debate about automation in professional services, with bar associations in three states already convening emergency meetings.
What the Model Can and Cannot Do
OpenMind was careful to frame the result in context. LexAI-4 excels at statutory analysis, case law retrieval, and contract review, but still struggles with courtroom advocacy, client empathy, and novel ethical dilemmas that fall outside its training distribution.
"This is a tool, not a lawyer," said OpenMind CEO Priya Nair. "But we expect it to fundamentally change how legal work gets done — in the same way calculators changed accounting."
Law school enrollment figures are already trending down for the third consecutive year. Legal aid organisations, however, see an opportunity: providing affordable legal representation to the 60% of Americans who currently cannot afford an attorney.