Karl McCollester (00:02.086)
Welcome to Mighty Municipalities, the podcast about small municipalities and how they can punch above their weight. I'm your host, Karl McAllister, and along with my co-host, Mark Parton, and our guests, we take you through the highs and lows of working in a small municipality. Whether it's the hardship and frustration that is holding you back or the progress and winds that are propelling you forward, we're here to share the lessons learned and give you strategies to help improve your community. Mark, how are you doing this Fourth of July weekend?

Mark Partin (00:29.883)
Doing great. It's been a a nice weekend and been a hot weekend in South Carolina but but had a good time and hard to believe we're celebrating two hundred and fifty years of our nation's founding. So it's it's a big weekend.

Karl McCollester (00:35.906)
Yes.

Karl McCollester (00:45.578)
It is, it is. And hopefully we can do two fifty more. you know. At least start catching up proportionally to some of those other other countries out there. So well great. Well today, you know, we were of course we were working on some projects and this one kinda this topic kind of landed in our lap and and I think it's one that

Mark Partin (00:49.607)
That's right. Yeah.

Mark Partin (00:53.519)
Yeah.

Karl McCollester (01:09.152)
A lot of towns are discussing and trying to figure out what to do with and that is AI. And Mark, you wanna give a little more about how this fell on our lap?

Mark Partin (01:21.603)
So as Karl mentioned, a lot of towns as well as organizations and people are talking a lot about AI because it is just at the forefront of technology as well as business in general. And with it comes some great opportunities, some great efficiencies and tools, but also some challenges and some things to be cautious of. And that's what brought us here to discuss this today.

And particularly if we're talking about small towns and and towns of all sizes, we have to think about the protection of data, the exposure that is is given when we put information out there or when we use these tools. But then also we have to think about the legal ramifications and requirements placed on governments, such as, you know, we're supposed to

store data in the government cloud which is loc located inside the US as opposed to servers that are all around the world. And you know the internet is designed so that servers are located everywhere for redundancy, efficiency, etc. But when you're talking about local government as well as state and federal government data, it needs to be located in the government cloud. And you know if employees fairly enough

don't understand that, they may think, well, why can't I not go out and just use, you know, one of the commercially available models. And so they go out and start using it, putting uploading information, you know, running analysis on it or writing letters, et cetera. And then all of a sudden confidential information is out there. And once it's out there, it's out there because that's what helps train these large language models.

Karl McCollester (03:03.479)
Mm-hmm.

Mark Partin (03:15.867)
become more and more efficient. There's more and more information so that it is able to work better. But if the the data is legally protected either because it's personally identifiable information or health information or just confidential intelligence, for instance, once it's out there, it's out there. And so we need to be aware of how to properly use it, which models

Karl McCollester (03:35.383)
Yeah.

Mark Partin (03:45.031)
to properly use, you know, which business associate agreements, for lack of a better term, define those scopes and are okay to use. And that brings us where we are today because if employees and citizens don't have the knowledge, you can't blame them. So we we need to educate people on what is appropriate and what is not. So for as part of our conversation today, Karl, could you talk through us a little bit about

what is AI and and explain a little bit how these models work for people.

Karl McCollester (04:19.255)
Sure. and so for AI can mean a lot of things. if you think about you know, s the phone calls we've been getting for probably going on what five years now, that has some sort of ability to understand your speech and respond to it. That's a version of AI. if you use Google Photos and you notice that it figures out your it groups all the

pictures of the same person together and helps you figure, you helps then say, hey, who is this person so they can then label it for you later. That's a version of AI. But for today, what we're gonna be talking about is lar large language models. And you know, of course we I'm sure most of you know exactly, you know, what that is given the the hype and the excitement around it. But just to kind of give a simplified version of what a large language model is, Mark, you're gonna be my large language model for today.

Mark Partin (05:18.599)
Okay.

Karl McCollester (05:19.071)
So, Mark, if I say to channel some of the World Cup hype, if I say take me home country roads, what how what how do you respond? What's the next part? Very nice, very nice. Okay, okay, very good, good, yeah, good, yeah, exactly. Well, so so does an LLM Okay, and if I said something like

Mark Partin (05:35.645)
to a place where I belong. I had to think a minute.

Karl McCollester (05:47.238)
I am the resurrection and the life.

Karl McCollester (05:52.672)
As a Bible verse. Okay. Yeah, that's pretty good. That's pretty good. Okay, so so great and congratulations, Mark. You are now a large language model. Because that is basically what a large language model does. it takes the information that we give it, and then based on that information, is finding at every word what is the next best word. Now imagine instead of Mark

Mark Partin (05:52.795)
No one comes unto the Father but by me. Yeah. Yeah.

Mark Partin (06:00.806)
Yes.

Karl McCollester (06:17.013)
happening to know you know, kinda these two s these two songs or the song and this Bible verse, let's imagine Mark had actually the access to every version of the Bible that's on the internet, every song that's lyric that's on the internet, all the chats in and reddit in the world that talked about these songs and about those Bible verses, every free article that could get its hands on about these two things. then, in addition to that, you know, Mark could have

Mark Partin (06:20.551)
Yeah.

Karl McCollester (06:45.597)
added to the that the rest of the thing or talked to then about John Denver or then talked about what book and what verse that was and and all of those things and because it would it would be able to basically take all that information and find the next part of it and and and can and continue to build more and more sentences about it. And that's basically how a large language model works. It's take it has all this knowledge

And kind of like Mark's brain and our brain, when it sees something or sees a question, it's basically pulling what it feels like is the next best word to complete that

that response. so it's not you know necessarily creating new things per se, except in the idea that of course if it's pulling from two different sources that haven't been pulled before, it may seem new to us. but it really does pull from everything that it knows. It just happens to be very good at pulling from you know thousands of sources or

For the rest of us, you know, for us mere humans in our brain, you know, we may be pulling intuitively at, you know, hundreds of sources and as well as some amalgamation of things that we've learned over time. So that's the basics of of what a large language model is and how those ch things like chat work. But chat isn't the only thing that it gets used for, Mark, is it?

Mark Partin (08:09.323)
absolutely not. chat is used in in several AI and chat is used in several different applications. And some of the more popular, well known ones are, you know, Copilot and ChatGPT, Claude, Gemini, those different models. But they're used a lot. sometimes it's annoying. You go on a website and it'll pop up and say, Hi, this is and it'll be some catchy name. How can I help you today?

Karl McCollester (08:32.311)
Mm-hmm.

Mark Partin (08:38.647)
And more times than not now, that's not a human there. That is a a an AI bot that has been placed on a website to help direct you either to answer a question or a query to make your your visit to that site goes more smoothly. Instead of you just having to search through all the different links, it'll take you to that one particular area. So that's that's one way it's used. but then also you can ask

a a large language model or a chat to to help you analyze or process information or data. You know, you can you feed in some numbers and say, what trends do you see here? And it it very efficiently will produce a a result. And you know of course you need to double check it and have a general idea of what the result should be, but it does make it more efficient and fast processing wise.

But then also, you know, you mentioned phone calls earlier and and matching up photos and so forth. But one thing that that AI models are really good at is if you're on Teams or Zoom or or you know any of those meeting websites, they're really good at taking notes and transcribing the conversation that's going on and then following it, they'll produce a a summary and in some cases if you've got it set up right, they'll assign

task to each of the different participants of the meeting. So all of those are really good useful tools that AI is currently doing for us and you know we don't want to negate those or minimize the effectiveness of that. But then you can also in some creative ways you can throw in some prompts and say draft me a a business letter on such and such topic

Karl McCollester (10:20.802)
Mm-hmm.

Mark Partin (10:32.453)
And it'll give you a good outline that again you should go back and put it in your own voice and make sure everything is correct, but it can get you good a good head start on the the document itself. but then also particularly for our finance friends and maybe some other bookkeeping friends, you may have some questions about what does this expense

look like or where should it be charged or or how should I charge this? And you can ask. You can say, you know, I spent money on this thing, this much money, this dollar amount, where's the best place to charge it based on my chart of accounts? And it will give you some good suggestions. So those are just some really simple helpful ways that that AI can make our lives easier and more efficient.

but again we do have to be the overseer of it and make sure that it is doing it appropriately and correctly.

Karl McCollester (11:37.742)
Right.

And I I think the point to that also is that, you know, it's not just chat, right? It is everywhere. you know, if you're you know, when we talk about the you know, as I'm about to talk about the risks, keep in mind that it's not just thinking about, I'm not using chat. If you're using you know, one of these like cloud desktop or even co pilot built into Word, you're using AI. If you're using one of those meeting recorders and summarizations, that is

is AI and that that represents the same risks. And really those risks are that for most of the tools, especially ones that do not have a government specific program or an enterprise program or option, and or or if you're not paying for that, most importantly, part of their

contract with you is that they get to use the data that you're putting in for training for themselves. And this makes perfect sense, right? If you're asking it to do to write a letter and then you're going back and forth with it three or four times about how this part isn't right and they need better tone and stuff like that. Well all of that gets fed back into the to the chat system so that it's performs better next time. Same thing if you're giving it any feedback on like how many stars if they did a transcription, you know, did it get every

And then you say, hey, it it missed this part, this part, and then didn't create the task for this. Again, it's gonna then use both that feedback along with the voice file recording itself to make that better. So why is that a problem? Well, since it's been used for that training, the thing is that just like

Karl McCollester (13:24.009)
Mark remembering country roads, Mark probably remembers somewhere like about an HR incident w about where he works, or Mark also remembers because those that is information he's received. So his brain has used that in training. And the same thing happens with with chatbots or with these AI based programs. If they're in reusing the information you've given it for training, now all of a sudden that knowledge, so you know, if you've had a phone call w that

Know you the department had their regular update, and then at the end you and your assistant were talking then about some you know job performance issues that people are having. Now that job performance information and that person's name that you mentioned during that that meeting is now in the you know qu in the in the data that that LLM is using for training. And so what they have found and you were flagging some.

things from the MIT AI Risk Initiative database, which is basically a place where people can track and record what things have happened with AI. That have re represent a security risk or a data leakage or things like that. And what

People have found especially of course people who want to find these things, hackers and so forth, have found out that if they persistently try different ways to ask these pieces of software about things that they think you know might be in there. So, you know, hey,

I'm really curious about the you know HR standards. I like to know what other municipalities have done. Can you give me specific examples? things like that, which may, by seeming innocuous, bypass the safeguards of the of the chat interfaces, or just because they have that data there. And now that data, again, that conversation is in a voice file and is also in a text file where it got transcribed, is sitting there on a on a hard

Karl McCollester (15:31.368)
Drive somewhere in that company's ecosystem, all of these things become at risk to be to be stolen and to be used. And that is not something that's most famous, most of the famous examples have been less recent, but even as recently, for example, as December, ChatGPT, since they were doing some human review pieces, their contractors were

actually then reusing some of the data that they were seeing as they were you know helping chat chat gpt reuse pieces. and so as a result some data ex got exposed specifically from a housing program. So that is a very government based one. Back in January there was a leak of private conversations that happened that were being stored within the product called Mooltbook And then in March

the Meta and their LLM and their data that's doing things around their glasses and stuff like that, they didn't have safeguards in place well enough so that unauthorized employees were getting access to find things out about other people within within that organization. And again that seems

like we've you know, we've known and we've hit all these things with things like using Facebook and things like using Twitter, but now we're giving them a lot more data and we're potentially using them for privacy, for things that should be private within the town. And so there's this exposure that we as municipalities have that we need to try to mitigate and we need to try to make sure our users and our employees are prepared and are guarded enough and make sure that they keep those that data safe. And so

Yeah, there's several on the good on the bright side, you know, so there's all that. there's a there is a lot of risk out there. but as Mark said, there's a lot of things that these systems can do. And if you've used them and you've started to rely on them, you realize you really like using them because they speed up your time or you know, that you don't have to think as hard about writing a policy or writing an ordinance because it's really good about helping generate those things. So what can we do to make sure that we're s

Karl McCollester (17:41.085)
able to leverage these types of tools while minimizing the risk to our municipality.

Mark Partin (17:48.946)
Well, there's there's several things that we can and should be doing and not just for AI, but you know, all smart systems, IT systems in general. one you you we need to educate our people through appropriate staff training. And just like we're talking about how these systems are used and how they work and some of the the risk of them on this podcast, this is educational. So one people need to know because if all they

have knowledge of is what they see on advertisements. They th this is cool, let me, you know, go draft this email or l or let the the model draft the email for me and it's great. but they don't know what is really happening behind the scenes and you know they're you know you they've been exposed and had had information exposed and and released. So staff training is is key. One to understand at a basic level

Karl McCollester (18:40.589)
Mm-hmm.

Mark Partin (18:47.941)
what LLMs are, how they work, how they retain and use data, and then also what is acceptable to do, what what uses are acceptable, which platforms are acceptable. Because if if we don't train people how to properly use it, we can't blame them if they misuse it. So s staff training is very, very important, particularly in government settings where there are legal requirements that

you know, may go beyond what happens in the in the business environment. But then also you have to be approaching it from a security point of view, which again if we're since we're talking in the IT world today, cyber security is just a huge topic that, you know, budgets for cybersecurity continue to grow, grow, grow because the opportunities for incidents continues to grow, grow, grow and become more complex. So

We need to make sure that when we integrate these systems into our overall platforms that we make sure that the right access limits are put in place for the user as well as for the the LLM itself. you know, if they should not be accessing HR files, then make sure that those are blocked and access is not granted. And same thing with you know contracts or whatever the the case may be.

Test it before you put it in production. You know, give a someone a chance to try to access it and make sure they're not able to unless they are authorized to. So that is one way. But then also go beyond that and make sure that people, again, through the training, understand we do not put in PII information or health information, you know, ju you just don't enter it.

And if you do need to ask some question around an HR instance, it's generic. It's not specific. You know, you know, like can you explain what harassment means or bullying means, but not give specific exempt examples of it. don't give access to those items that people wouldn't have access to access to without a large language model.

Mark Partin (21:08.945)
So if someone doesn't need access to HR, then they the system doesn't need access to HR when they're using it. You know, so just some common sense things like that. And then we need to be make sure we're being very cautious when we select different products, different software products, and I know you're going to talk about this in a minute, but try to be aware of what is going on in the background of these packages that

they are accessing and what they are processing and doing to make sure that they aren't that we're not ignorant to it, you know, that that they're not doing something that we wouldn't want them to do, but we're not directly involved in how it was set up. And you know, camera systems are one of those things because they do have great AI capabilities, but we need to make sure they're not processing something they should not be processing just because it was in view of the camera. So

Karl McCollester (21:55.693)
Mm-hmm.

Karl McCollester (22:08.429)
Absolutely.

Mark Partin (22:08.699)
So can you tell us a little bit about maybe some of the governance structures around how we approach setting up an AI system in in a local government? Yeah.

Karl McCollester (22:18.551)
Sure, yeah. So as Mark mentioned, when we're talking about the security, right, the dangers are, you know not only not including PII, which is priv privately identifiable information or personally identifiable information. There you go. which is things like, you know, people's names, addresses, social security numbers, all that kind of information. Information that could be used to identify the person that we're talking about.

Those kind of things, we're also then of course want to be very careful about what

Files we give it access to that may have that very same information. So how do we figure out what it's what it's using? Well, you know, unfortunately, and this is true for small communities as well as large communities, we need to have a good understanding of what tools are currently being used, and then making sure that those are being used and or can be managed in a way so that they are used safely. So what does that mean? whether

In general, what that first starts to is whether of all software, whether that's AI enabled software, AI chatbots, even your finance software, even your you know, office you know, productivity software, understanding and keeping a list of you know what tools are of those are being used in in the environment, and what AI policies do they have and what can they access.

And how much of that access do you have control over then what goes up into the into the larger chat bot back for training, right? So to give kind of give you some examples of that, so of course w you know most of us use Microsoft Office and but the same thing's true for Google Drive or Google Suite of products. they both have government programs that limit what data

Karl McCollester (24:21.057)
that they see actually get shared back with the model for training. So if you're in a government program with Microsoft or with Google, you have some protections built in that keep that data from leaving your your world. The large language model can use it to help you get an answer, but they are their contract terms keep it from then being used to go back and train the model.

So that gives you a measure of protection. But you have to make sure that you're on the government program. So if you just went to Microsoft.com and bought it, you don't have those protections. So you need to make sure that you're on the right program. So, you know, going through, so just you know, starting with Office, going through what chatbots y'all are using, you know, again, looking at you know what exactly is QuickBooks looking at and using and trying to figure out what my data protections are there if I'm using QuickBooks.

So then that's so now that we have that list of software again going back through each of those pieces and looking at what does the license protect for me. And if the license doesn't protect me, then under then making sure that our folks understand you know this tool, if we put information in here that's identifiable or to of about our citizens or about our employees, this tool

Could be a security risk. So in those cases, either, you know, that's why we're not going to use this tool, or if we do use this tool, w we need to make sure that we're not sharing that type of information. So updating, you know, having that list of software, reviewing it, sharing it with everybody regularly or putting in some place where everybody can see. So if they're like, hey,

I'm wondering I'm thinking about using this software, which ones can I use? or you know, hey, I I got this new software, how do we figure out whether we can use it or not? Again, having that procedure in place. And I know for a lot of our smallest towns, this can be difficult, but hey, this could just be an Excel spreadsheet that that you keep in one place, or it's a very much a, you know, it's on a whiteboard and you're like, hey, if you need a new piece of software, come in, let's add it to the list, let's make sure we let's go look at it together and understand exactly what

Karl McCollester (26:37.835)
we could do and you know you can always start very carefully right again you could say okay we can use it don't share anything with it and be very clear about that and then going back and even at that for those older pieces of software that we've reviewed once making sure you're going back and reviewing them annually because things are changing so fast there's new features being added all the time you know Microsoft Office didn't have

half the features two years ago of what it has today as far as how it interacts with large language model with co-pilot and does all those things. So making sure that that those things are reviewed and kept up to date. And if it becomes a risk, making sure you're taking it off the the the board of what things are available. So that's the general so first again that first part is figure out where you're going to maintain that list of what software is allowed by the the municipality and on the other hand what is not.

You know, again, I'd be very careful, you know, about you using commercial chatbots. And so it may be that, you know, hey, these things are allowed for you know, these couple things are allowed for anything. Maybe you decide you trust Microsoft because enough that they already have all your files, or Google Drive and they already have all your files. They are you know, and so they've already got that data. The license that you have, you're in a government program, they promise not to share it, so everything's okay. And then the next tier may be.

the you know we don't have that with chat GPT but we f or you know we have and we don't have that with Grok but we feel like those two are trustworthy enough that we should or productive enough for us that we should use but you know be very clear to everybody you cannot you know put anybody's name in it you can't put anybody's address in it all those kind of things and if you can't ask the question without doing it then don't ask the question.

And then finally, here's the list of things that, you know, yes, there's this great new tool that does talks to five different chats all the time and you know does all these aggregation things, or there's this great new meeting software that's going to record everything and put it in, but we just don't know about it. These things are not allowed yet. So that's that first piece, that software inventory.

Karl McCollester (28:48.588)
The next piece is then making sure your employees understand how they're using it and also that just because it's AI does not change their responsibility. you know, Mark mentioned training earlier and yes, figuring out how to get that training one is a pain and two right now is

very up in the air. There's there are pieces of things out there on the internet. You can feel free to use this podcast. But there are lots of things that you know there it's very hard to find definitive training right now. I certainly people are working on that. But that's the same thing that's true for our police officers for CJIS So if you've if you have a police force you already have this kind of responsibility, it's just extending it to the rest of your employees and figuring out what you're gonna do. Again, whether it's listening to this, whether it's giving them a handout, whether it's having an annual

Mark Partin (29:14.513)
Yes.

Karl McCollester (29:37.825)
Annual you know all hands meeting where everybody comes in and you just talk through these things or have your your IT person talk through these things or your IT managed service provider talk through these things, just so everybody's on the same page. but again, employees you want to make sure your employees understand that.

Just because it's AI, they are still responsible for what they pull out of that. So if they write a letter and the AI makes a mistake, that's not the AI's mistake. That is their mistake and they're held responsible for it. If they leak data and give data to you know the AI and the AI and then that is discovered by somebody else, that is not the AI's fault. That is in the end their responsibility because they they should not have been sharing it if it was not in your policy of one of the tools that was okay to share. So it is very important under for

you to make sure your employees understand that AI is not a get out of jail free or a hall pass that lets them take any action because it makes them more productive. So those you know so again first figure out what your software is, two making sure your employees understand how to use it. And then finally three figuring out what are we going to do when something happens.

What we do if something happens during a security issue? Right? If hackers came in and got all our files locally and encrypted them, what do we do? and hopefully you already have a plan for that. Well, this is just adding some pieces around what do we do if we discover that we've shared information that should not have been shared with an AI, right? It's it's close to a hacker. It's hopefully less

destructive to the municipality than a hacker. But it's still something you may want to acknowledge or you know that I would encourage you to acknowledge. I would encourage you to understand how you're going to respond to that. because it is very close to that same sort of issue. So figuring out how you're gonna respond to that and having that information, having that plan ready in case and if and when that does happen, along with your security plan or as part of your security plan, certainly helps you

Mark Partin (31:22.898)
Mm-hmm.

Karl McCollester (31:51.154)
be able to then respond and also helps employees understand of why it's important to report that and why and and what happens when those kind of things occur.

Mark Partin (32:01.655)
to to go along with what you were saying about vetting some of these tools, it just reminded me that you know for this collection of data is nothing new. It's been going on, you know, forever and ever and ever, and even before AI became a a common language that we use. But anytime you sign up on a website as well as it you know used to be in person,

They'll tell you to read the customer service agreement and the privacy policy. And most people just click through it and go on. But when you're talking about tools like this, take the time to read it and and you they'll they'll say what they're going to do with it and what they're not going to do with it. so that is an easy first step you can take to filter out and understand what is happening with the data that you're giving them because the data is is the

the valuable asset that they have because without data they can't do anything. These systems cannot do anything. It just it requires data. so read those customer service and privacy agreements and you'll have a a better understanding. Hopefully not too shocking of one, but you will have a an understanding of it. Yeah. So

Karl McCollester (33:06.808)
Yeah. Yes.

Karl McCollester (33:20.608)
I and you know, just like anything else that comes in over the internet, treat it like treat that software

like something to be suspectful of. Yes, it's been great answering your questions for the last two months, but if it's suddenly asking you or giving you the opportunity to share all your local files with it to, you know, and I'm using air quotes here that you can't see, make the make the system better or make it work for you better, pause and think about what might be in those files. You know, you wouldn't share those by default with any other program that you hadn't licensed, especially a free program.

or a program that you're only paying thirty bucks a month for without understanding how is that going to be used and where is that going to be stored and how do I get it back. So the same thing's true with with an LLM. If you don't understand

where it's going to go, you want to be very, very careful about what you're providing to it. because it's very easy to accidentally give it a file that you did not mean to, or access to some subfolder or some full subfolder that all of a sudden you've now provided employee information or information about utility billing subscribers and and or businesses and things like that.

Mark Partin (34:36.413)
Yes. All right, Karl. So now I think it's time to move on to the discussion of grants this week for the podcast. So the first one we'd like to report on is is due this coming week very soon, so two days from now actually. so hopefully you either already knew about it or you have time to to get ready for next year. But it's a grant by the National Park Service for historic preservation.

Karl McCollester (34:52.046)
Yes. Yeah.

Karl McCollester (35:00.494)
Yeah.

Mark Partin (35:05.369)
And it's it's particularly focused on underrepresented communities and a really nice grant a total of fifteen thousand to a hundred thousand dollars and it's to help ID historic sites and help get them on the National Register of Historic Places.

Karl McCollester (35:23.596)
Right, and only slightly less hairy of a deadline. the Health Health Resources and Services Administration, which I didn't know existed until I got this, the HRSA has a rural communities opioid response program. So that is up to 750,000 for four years to opioid treatment programs. so again, very short deadline, but you know, potentially something you can go for and or prepare for next year.

Mark Partin (35:51.134)
Alright, then we have another one that's due this week on Thursday. Yeah. That's right. Yeah. And this is by the National Endowment for the Arts. It's called the Grants for Arts Projects, and it is for art education and design, our town projects. And two different amounts here. The Challenge America program is a ten thousand dollar grant, and then some more general grants for general projects are

Karl McCollester (35:53.583)
You got one more day.

Mark Partin (36:21.155)
Anywhere from ten thousand to a hundred thousand dollars. So both of those are are two nice grants to look at from the National Endowment for the Arts.

Karl McCollester (36:29.536)
Yeah, and yeah, I looked I couldn't quite figure out what was what. I mean, just given by the grant size, I'm assuming the Challenge America program is much easier to get as a starting point, at the ten K and then they have more substantial grants from the ten to one hundred. So you know, go for it. you got you got three days from hopefully from the release of this podcast to get those ones. maybe you have enough time to get the Challenge America grant going. so then at least now we're up to having a week in advance. The consumer protection agency is has a pool safety program.

And those are grants of between 50,000 and 400,000 around pool safety education and enforcement. So if you, you know, you prototypically there are a lot of pools down here in the southeast. so this is about sharing that information about how to keep you know kids and pets safe and what to do around those pieces. And so implementing that sort of program in your municipality, and especially if you can if you've had a rash of those, where law enforcement

Forcome or EMS have had to come in and this gives you a way to potentially do something about that and reduce those in the future.

Mark Partin (37:37.266)
All right, the next grant we're going out a little bit to July twenty-ninth, so we're giving you a little bit more time. but yes. yeah, excuse me, July 13th. I skipped right over that. but this one actually r goes along with what we talked about on our last podcast and ways to to combat extreme heat. And this is for Georgia municipalities and it's called the Georgia Relief Grant, and that's spelled R E L E A F.

Karl McCollester (37:42.934)
I get one more on the thirteenth. Yeah. Yeah, right? Yeah, yeah, yeah.

Karl McCollester (37:55.629)
Yes.

Mark Partin (38:07.343)
Leaf as in Leaf on a Tree grant. And this is a grant for fifteen thousand dollars to plant trees in urban settings. And not only is it a beautification effort, but it it can also be used to fight against extreme heat. and if you did not listen to last the last podcast we had, encourage you to go back and listen to that and it matches up with this grant very nicely.

Karl McCollester (38:32.876)
Yep, and this is where I learned I I'd really need to s give better spacing for these for Mark and I when we're going through. But next up, there were a lot of grants coming up, so we've got two two more, second to last here. due at the end of the month, July twenty-ninth is the Department of Justice Cops hiring program. So if you are looking for funding to help hire additional law enforcement officers, this provides up to 125,000 over three years. It does require a twenty-five percent match.

Mark Partin (38:37.787)
Yeah.

Karl McCollester (39:02.85)
I think given today's cost of hiring law enforcement, that is gonna go quite quickly. but if you especially if you've have some amount, and I bet you're probably matching you know closer to 45 or 50 percent to make that 125k over three years work, you know, that is a great way to you know potentially offset some of those costs, especially if it's something where we're you know trying out a new program, like a cops on the streets program or maybe a school safety officer program in a new school.

or something like that to be able to reduce that cost.

Mark Partin (39:35.709)
And the final grant for today also can refer back to our podcast on fighting heat. It's for New York municipalities with a due date of July 31st. And this is the grants for climate action. And these grants, pretty big range of of awarded mounts here, $35,000 to $2 million for climate change mitigation, particularly around flood control and and renewable energy.

so tied also again to the heat and the the bigger issue of what causes heat and so forth, but nice grant there for New York municipalities to consider applying for in the next few weeks.

Karl McCollester (40:21.368)
Great. Well thanks Mark. you know, I think if I had one message for small municipalities with AI, AI is a great tool. it is absolutely something if

you can figure out how to use it and it's making you productive, it you shouldn't stop using it. Simply that you want to make sure you take the steps that you you're safely using it and even more importantly that your employees are safely using it because they may not have the same exposure or you know holistic view of the community. and so by laying out some of these policies

making sure that people understand what should and should not be used as part of their job really helps mitigate a lot of your risk and the risk of losing information you know out to bad actors unfortunately.

Mark Partin (41:15.161)
Absolutely. So a very timely and important topic for us to talk about today and I've enjoyed this conversation.

Karl McCollester (41:22.893)
Likewise, Mark. Thanks. Talk to you in a few weeks. Thanks. Bye.

Mark Partin (41:25.755)
All right, sounds good.