The AI was reviewed, benchmarked, and audited. What does that actually prove?
The Agentic Lawyer examines what AI reviews, benchmarks, audits, and vendor assurances can establish before lawyers rely on them.
Written by Lana Akopyan, patent attorney and former software engineer.
Subscribe and get AI Vendor Terms Triage, Field Guide No. 01.
New analysis by email, one piece at a time.
Author of The Asymmetry Audit, an evaluation of 13 commercial AI models
Article forthcoming, University of Florida Journal of Technology Law and Policy
IP Chair, Women's Bar Association of the State of New York
Published in IPWatchdog
Questions this publication works on
Does human review prove an AI workflow is reliable?
Does writing the first draft yourself preserve your judgment?
What does an AI evaluation actually establish?
Can last quarter's audit support today's system?
When does an accurate benchmark number mislead?
Why is a thorough AI draft a risk in a legal instrument?
The framework behind these questions
Evidence theater: when confidence outruns what the evidence can actually establish.
