Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Ai Policy topic
No spam. Unsubscribe anytime.
FPPC committee reviews draft generative AI use policy emphasizing human review and security
Summary
The FPPC's Public Outreach & Technology Committee reviewed a draft generative AI acceptable-use policy that requires human review of AI-generated work, restricts public-facing AI tools for now, and notes the agency's enterprise Microsoft Copilot deployment prevents prompts from training public models; the policy is expected to go to the full commission in March.
The California Fair Political Practices Commission's Public Outreach & Technology Committee on Feb. 11 reviewed a draft internal policy governing staff use of generative artificial intelligence, emphasizing human oversight, data security and compliance with public-records obligations.
General Counsel Dave Bainbridge summarized the policy, telling the committee that "when product is generated using AI, there is a human in the loop that reviews the content, and ensures its accuracy." The draft also identifies risks such as unauthorized data sharing, bias, overreliance and the disclosure of confidential information, and it includes a requirement that staff disclose to supervisors when they use generative AI as part of their work.
Security controls were a focal point: staff said the agency has procured an enterprise version of Microsoft Copilot that prevents prompts from being used to train public models and enables auditing of prompts if needed. "The enterprise version…keeps it everything in house, essentially," said Jesse Hidalgo, the commission's chief information officer, explaining the procurement choice and the distinction from public consumer accounts.
The policy, described as a first step, would require supervisor approval and IT review for any GenAI tools beyond the approved enterprise option and would mandate a disclaimer on public-facing products created in part with GenAI. Bainbridge said the draft will likely be presented to the full commission for approval in March, after which a final version will be posted publicly.
Committee members discussed coordination with the Department of Technology and a statewide AI community that shares guidance and best practices; staff said the department's security and privacy recommendations informed the draft. Chair Silver noted that while other jurisdictions have experimented with public-facing advice agents, the FPPC is not planning public-facing AI tools at this time because of the risks involved.
Next steps: staff will bring the draft to the full commission and continue staff training and working groups on GenAI use. The committee did not take a formal vote on the policy at this meeting; Bainbridge said a FAQ and training materials will accompany the policy when finalized.
The committee asked members of the regulated community and other agencies to monitor the policy as it evolves and to provide feedback through existing channels.
