Board Briefing Summary - February 2026 ====================================== Key regulations --------------- - Colorado SB 24-205 (AI Act) requires bias testing + opt-out controls for "high-risk" systems by June 30, 2026. - EU AI Act categorizes our discovery/chatbot stack as "high-risk". Vendors must provide technical documentation before August 2026. Action items ------------ 1. Contract amendments: add AI disclosure, deletion rights, 120-day notice for any model retraining that touches patron data. 2. Staff capacity: 20 hours/quarter for AI inventory + reporting; $12K placeholder for third-party bias testing shared across consortium. 3. Documentation: keep vendor assurance letters + testing logs for five years (public-record requests). Questions to ask vendors ------------------------ - How do you test for bias, and can we see the last report? - Where is patron data stored while models are trained? (Need U.S. residency.) - What is the opt-out workflow if a patron declines AI-generated suggestions? Contact: sam@unhingedlibrarian.com