About

Most guidance about AI in legal work arrives incomplete. Doctrine and ethics opinions describe duties without describing how these systems actually behave. Technical writeups describe behavior without touching the duty a lawyer owes when the output moves a file. This publication works the intersection.

I am a USPTO-registered patent attorney and a former software engineer. I build and evaluate AI-enabled legal systems and workflows, and I read the vendor terms governing the tools lawyers use.

The questions I bring to a tool are evaluation questions. What was the system actually tested for. What can the result establish. What evidence would survive reliance, if reliance became the issue.

Legal competence requires disciplined friction at the points where judgment matters. These systems are optimized for fluent assistance, not independent professional judgment. The famous failure, the invented citation, is the one the profession already catches. The expensive one is confident, correct-sounding output that quietly moves a decision the lawyer never made.

The pieces here name specific failure modes, cite primary sources, and read vendor terms closely. Material corrections are noted in the affected piece.

I serve as IP Chair of the Women's Bar Association of the State of New York. My article "Operational IP Debt: How AI-Driven Organizations Lose IP Value Before the Law Ever Applies" is forthcoming in the University of Florida Journal of Technology Law and Policy.

Educational only, not legal advice, and no attorney-client relationship is created. Views are my own. Attorney advertising in some jurisdictions.