HomeMy WebLinkAbout052026 email - AI policy feedbackALERT: BE CAUTIOUS This email originated outside the organization. Do not open attachments or click on links if you are not expecting them.
Commissioners,
Please refer to the attached, detailed AI-generated analysis of the county’s proposed AI policy, as it was documented in the 40-page agenda item in your May 18, 2026, meeting packet
(“WORKSHOP AI.pdf”).
I was disappointed in Mr. Frederick’s dismissive response to my public comment in which I referred to the presence of an implied NDA (Non-Disclosure Agreement).
Within the 3 minutes allotted, I did not say that there was a literal NDA in the policy. What I said was, “The other objectionable part is that there is a built-in unilateral NDA requirement
that says, you know, protecting county information is a responsibility of those [AI users]. And it says those responsibilities continue until the information is reclassified to be public.
Well, I didn't sign an NDA, and I won't.”
The proposed AI policy makes it perfectly clear that the policy will apply to all “AI users”, which definition includes volunteers such as myself.
Page 10: “AI User" means any person is using, developing, purchasing, configuring, or maintaining any AI used by or for the County, including contractors, consultants, elected officials,
employees, and volunteers.
So, yes, the acronym “NDA” does not appear in the policy; what is present in the policy is an NDA-like obligation that would be imposed on me should the policy be adopted as-written,
and with which I will not comply.
Page 24: AI users shall comply with all applicable County policies and are required to read and understand them. Responsibilities for protecting County information do not end at the
termination of employment. These responsibilities continue until the information is reclassified to be public.
“Protecting county information” is not my job. In my volunteer role as chair of the county’s Ferry Advisory Committee (FAC), I create a large number of public records (county information),
primarily in the form of emails. Because the county does not provide FAC members with county email addresses, we must rely on gmail or other 3rd party mail systems. While I am very
careful to use the gmail address that I have set up exclusively for FAC purposes, I cannot be held responsible if Google loses an email.
I’ll point out a couple of other problems in the policy which stand out as being impractical and/or unenforceable.
Page 15: Use of an AI system on a County-owned computer, server or system that is not approved pursuant to this AI Policy is prohibited.
Users of county-owned computers have no way of knowing which AI systems might be acting on their behalf as they access websites outside the county’s purview. A simple example is any
use of Facebook, where malevolent AI’s (almost certainly not on the “approved” list) lurk. Dozens of county staff are empowered and/or required as part of their jobs to monitor and
post on Facebook. Is the county going to take disciplinary action against them when they access a site that uses unapproved AI?
Page 40: Use of AI systems for purposes which may impact the administration or results of elections
The county administrator and staff have prepared a “fact sheet” for the upcoming parks and rec levy proposition; while such a fact sheet is nominally neutral, its existence (as is always
the case in such matters) cannot help but impact the decisions of the electorate.
Page 40: Use of AI tools which consume excessive energy or resources without consideration for environmental impacts
You cannot be serious! Users will have no way of knowing how much energy, excessive or not, their use of AI compute entails. “excessive” is meaningless.
Policies which are facially unenforceable should not be considered for adoption. The proposed AI policy is clearly flawed, and needs to go back to the drawing board. I would say, “Nice
try”, but that would not be honest.
Tom Thiersch