State technology officials brief committee on new AI policy, use cases and safeguards

Gretchen Perry, state chief technology officer, told the Technology, Economic Development and Veterans Committee that Washington must "strike a balance in the state between innovation and trust," calling trust "our currency" as agencies expand use of artificial intelligence in public services.

The briefing centered on the statewide AI policy WATEC adopted recently and governance steps designed to limit risks while enabling practical applications. "We have now, just just on December 11, adopted an AI policy for statewide," Katie Ruckel, the state's chief privacy officer, said, describing policy elements that mirror the NIST AI Risk Management Framework and add a "public purpose and social benefit" principle for state use.

Why it matters: officials said the policy requires annual application inventories where agencies identify AI use, a risk assessment process that differentiates low/moderate/high risk cases, privacy and security reviews for higher‑risk uses, accuracy checks and role‑based training. The state also has model contract terms and data‑sharing agreement (DSA) guidance to control how vendors use state data.

Agency use cases highlighted practical results and limits. James Galvin, Emerging Technology Program Manager, said the state's application inventory recorded "200 different AI applications" for largely low‑risk use cases. He described three examples: using satellite and geospatial data with AI to speed flood damage assessment and FEMA documentation after recent flooding; a Department of Licensing chatbot called "DALL E" that provides multilingual, self‑service responses; and a contract assistant limited to an approved knowledge base to help contract specialists find procedures and policy guidance.

Committee members pressed presenters on oversight and impacts. Chair Cindy Ryu asked whether the contract assistant accounted for audit needs; Galvin said the tool is constrained to the desk manual and is not configured for auditing. Several members asked whether conversational translation efforts include professional interpreters for court and health care settings; Galvin said language services are a broader working group and that work to date has focused on chatbots and call‑center support, not interpreter substitution.

Labor and vendor oversight were recurring themes. Ruckel said agencies follow an Office of Financial Management directive requiring six months' notice to unions when AI would have a consequential effect on hours, wages or working conditions, and that unions have been notified and in some cases bargained over policy language. On vendor choices, Galvin said the "vast majority" of the state's deployments use off‑the‑shelf large language models and that bespoke models are reserved for niche cases where custom training is justified.

What comes next: presenters said the policy is intended to encourage safe adoption by reducing burden for low and moderate risk use cases while requiring deeper review for high‑risk projects. Committee members were invited to review the policy online and to continue consultation as agencies pilot and scale tools. The committee adjourned without taking formal action on any bill.