Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Technology Policy topic
No spam. Unsubscribe anytime.
FPPC adopts generative-AI acceptable-use policy requiring human review and training
Summary
The Fair Political Practices Commission approved a policy governing staff use of generative AI, requiring human-in-the-loop review, disclosure to supervisors when AI is used for work products, enterprise-level tools and mandatory training before staff may use the tools on agency work.
The Fair Political Practices Commission voted on March 19 to adopt a generative-AI acceptable-use policy that sets guardrails for staff use of large language models and other generative tools. The policy emphasizes three core controls: human-in-the-loop review, use of approved enterprise platforms, and mandatory training.
What the policy requires: Executive Director West and IT staff explained that staff must keep a human in the loop — outputs from generative AI are the user's work product and must be reviewed and verified before release. The policy restricts staff to approved enterprise tools (the commission identified Microsoft Copilot as an approved vendor platform) so prompts and outputs are retained within agency-controlled systems and not used to train third-party models. Staff must notify supervisors when generative AI is used for work product and a public-facing disclaimer is required when AI materially contributed to published content. Mandatory training and an early-adopter pilot will guide phased rollouts.
Reasoning and safeguards: CIO Jesse Hidalgo and Executive Director West said enterprise software, review requirements, supervisory disclosure and training mitigate primary risks — inadvertent disclosure of confidential information, hallucinated outputs, and vendor data-sharing — while allowing staff to gain efficiency on routine tasks. West summarized the human-in-the-loop principle: "Employees using generative AI — it's still your work product; you are responsible for it and must review any outputs."
Adoption and oversight: Commissioners discussed internal controls, auditing prompts, and how supervisory checks will work in practice; the policy passed on a commission vote that recorded unanimous support. Staff will run a small pilot, provide training, monitor use, and return to the commission with updates and FAQs.
Next steps: implement required training for staff volunteers, pilot approved enterprise services, collect employee feedback and report to the commission on early use cases and any necessary adjustments.
