Managing AI Legal Risks: An In-House Counsel's Practical Guide

Abstract

A contract for an AI-powered hiring tool came across my desk for review. The agreement resembled every other software license we had reviewed, complete with standard warranties, intellectual property indemnification, and limitation of liability provisions. However, buried in the technical specifications was a detail that fundamentally altered the risk profile. The AI system had been trained on millions of resumes and job descriptions collected from publicly available sources. That single phrase represented potential exposure to copyright infringement claims, discrimination lawsuits, and regulatory enforcement actions that our standard contract terms never contemplated. The vendor's indemnification clause covered intellectual property infringement but explicitly excluded claims related to training data or AI-generated outputs. We were about to inherit liability for a system trained on data of unknown provenance, with no legal protection if things went wrong.