Key findings
AI is most valuable for structuring and organizing. In legal work, this translates to tasks such as drafting checklists or categorizing contract clauses.
Major quality gaps between systems. Where other AIs make errors more often, the paid versions of ChatGPT 4o Plus and Claude 3.5 Sonnet perform best.
Critical thinking beyond blind faith. AI systems sometimes produce legal information that sounds convincing but is actually incorrect. Critical thinking and verification remain essential.
Compliance remains a point of attention. Although some AI systems offer good safeguards for confidentiality, significant legal risks remain.