[Chad Davis] 12:04:10
Hey everybody, thanks for joining us for our July Public Media Innovators webinar. I’m Chad Davis, co-founder of Public Media Innovators, also Chief Innovation Strategist at Nebraska Public Media. I also want to introduce my
[Chad Davis] 12:04:25
co-founder Amber Samdahl. You want to say hi?
[Amber Samdahl] 12:04:29
Hi, everybody. My name is Amber Samdahl. I’m Director of Design & Innovation at PBS Wisconsin, and excited to have you all join today.
[Chad Davis] 12:04:39
Cool. Amber’s gonna post some links for this particular webinar, since it’s policy-based, we are…
[Chad Davis] 12:04:48
We have collected some policies on our Public Media Innovators website, so you can check those out. They’re there, and we thank everyone who shared theirs with us so that we could kind of create some resources for y’all to draw on as…
[Chad Davis] 12:05:04
we’re doing this particular webinar and for you to use after the fact. A little bit of housekeeping before we get into this. Next webinar is going to be, I think it’s August 20th.
[Chad Davis] 12:05:19
Third Thursday, as we like to do. We are going to be hosting someone from the think tank New Public. They’ve recently published a deck that looks at sort of the life of people online after sort of the…
[Chad Davis] 12:05:36
social feed is supplanted by AI, it’s called After the Feed, but they’re gonna give us a special kind of public media tuned version of that, so, uh, should be able to, to, uh, be able to check that out. Amber just put the, um, the link in, calling it AI and the Future of Community. So that’s coming up next month, join us for that.
[Chad Davis] 12:05:56
And then today is like, we have an embarrassment of riches because at the top of the next hour, PBS is actually doing a webinar on AI policy, and it’s coming out of their standards and practices.
[Chad Davis] 12:06:10
Uh, team. So we’re gonna try to wrap this up just a little bit before the top of the hour so that we can get you all over there. And what we would love to do, uh.
[Chad Davis] 12:06:18
um, is a Twitch-style raid on the PBS webinar, if you are Twitch people.
[Chad Davis] 12:06:25
there’s this function where you can actually, like, send traffic over to a different stream once your stream is done. We can’t do that here, and that wouldn’t be very public media, even if we could. But we’re gonna post a link here in a second, and…
[Chad Davis] 12:06:40
We will encourage you to jump over and do PBS’s webinar.
[Chad Davis] 12:06:47
At the top of this, when you get there, tell them Public Media Innovators has sent you. We’ll put some link in, we’ll remind you at the end, but we think this could be the first webinar-to-webinar raid in public media history, and we love our firsts here at Public Media Innovators. Thanks to NETA, for always supporting us in doing these things.
[Chad Davis] 12:07:05
And crazy ideas like webinar to webinar raids. And thanks to Current, who collaborates with us quarterly on certain webinars, and they came back around for this one on AI policy. We are about to launch a little poll to get you all started.
[Chad Davis] 12:07:20
by way of a bit of an icebreaker. And while that is going in there, I’m gonna toss it over to Mike Janssen, our collaborator over at Current. Mike, over to you.
[Amber Samdahl] 12:07:38
You’re muted, Mike.
[Mike Janssen] 12:07:41
Sorry about that. Yes, hi, thank you. And yes, please take this poll so we can get a bit of a pulse on where everyone is with AI policies at their stations. So I’m Mike Janssen, digital editor for Current, and thank you to all of us for joining us today. Really great to have you.
[Mike Janssen] 12:07:58
No matter what your station or organization is doing with AI, it’s becoming increasingly important to have an AI policy.
[Mike Janssen] 12:08:05
Employees need guidance on how to use AI in ways that protect trust, accuracy, and your organization’s reputation. And as media consumers become more wary of low-quality AI-generated content, audiences need assurance that the information you’re providing has been vetted and shaped by people.
[Mike Janssen] 12:08:22
Today, we’re going to talk with representatives from three, hopefully four, stations, if we can get our fourth panelist in here, that have developed policies using very different approaches that reflect where they are in their adoption of AI.
[Mike Janssen] 12:08:36
Actually, I’m just checking. Do we actually have — yes. We’re being joined by Drew Brody. Thank you. Glad we have him. All right. So we do have four stations represented. We have Drew Brody just joining. He is the chief operating officer at Chicago Public Media.
[Mike Janssen] 12:08:51
Margaux Maxwell, Director of Platform and Product for Oregon Public Broadcasting. Maggie Ryan, Contributing Editor for Little Rock Public Radio in Arkansas.
[Mike Janssen] 12:09:01
And Kyle Vernier, senior managing producer at Detroit PBS.
[Mike Janssen] 12:09:06
So before we start the discussion, let’s have a look at the poll results.
[Mike Janssen] 12:09:11
Where are stations in developing AI policies?
[Mike Janssen] 12:09:14
It’s pretty all over the place. Interesting. About 21% say they haven’t started. 21% exploring whether to create one. OK, interesting. Good to see that. All right.
[Mike Janssen] 12:09:29
So let’s get started with our conversation. Let’s start with Maggie at Little Rock Public Radio. Your station moved quickly to develop a policy once your staff started talking about using AI. What made you feel that the newsroom needed a policy before experimentation with AI went any further?
[Maggie Ryan] 12:09:48
Hey, Mike, thanks. So first, we really wanted to get ahead of these systems before they became more widely adopted. And we wanted to be very clear about how we were going to use AI if we were going to use AI as a newsroom. We were a very small newsroom at the time. We had four people in the newsroom. So we really wanted to do this in a way that.
[Maggie Ryan] 12:10:09
one kept everybody on the same page about how the technology was being used, and also maintained this relationship of trust and transparency that we have with our audience as a public radio station, and as one of the larger public radio stations in our state. We didn’t want to do anything or risk doing anything that would jeopardize that relationship of trust and transparency with.
[Maggie Ryan] 12:10:30
our listeners or with other newsrooms that we collaborate with.
[Mike Janssen] 12:10:36
Drew, what prompted Chicago Public Media to create a policy?
[Drew Brody] 12:10:40
Yeah, when I first came to Chicago Public Media in November of 2024, there was no policy at all. There was one line in our handbook that said, if you are interested in AI tools, talk to your supervisor.
[Drew Brody] 12:10:55
And this was getting to be a problem, because our supervisors would come to leadership and say, what am I supposed to say here? So there was… we were sort of in a doom loop of no answers. And it was getting clear that we needed to have a policy that would.
[Drew Brody] 12:11:12
protect… that would… that would both be audience-facing, so that we could, uh, ensure… help our audience understand where and… where we do use AI, and where we would draw the line. Um, and also to have the internal conversation. We just hadn’t had these talks as, uh, as an organization about.
[Drew Brody] 12:11:30
how everyone feels about AI, uh, where they think that line should be in their work. People were starting to experiment in little pockets, um, I would say mostly on the business side, and less so on the editorial side. Um, uh, and we… we felt like we needed really.
[Drew Brody] 12:11:45
to have the conversation, and then to have really clear guidelines about, um, the work that we would do involving AI, so that we could be super honest and super transparent with our audience.
[Mike Janssen] 12:11:57
Kyle and Margaux, were your reasons similar to those, or did you have a different trigger? Kyle, why don’t you go first?
[Kyle Vernier] 12:12:04
Uh, yeah, I mean, our reasons are very, very similar, you know. Um, I think we, we had staff that you could kind of tell we’re, you know, experimenting with it, but we didn’t really have, like, a full understanding of, like, what people were doing, and, and we wanted, uh, other policies were starting to come out around the time we, we started, decided to develop ours.
[Kyle Vernier] 12:12:22
And so we, you know, we thought, well, this would be a good way to help guide what people, you know, people I think had questions about what is appropriate use of it, you know, and so creating a policy, I think, helped us inform the staff, like, about the ways that they can use it.
[Mike Janssen] 12:12:39
Margaux?
[Margaux Maxwell] 12:12:40
Yeah, first of all, I would say we’re in a really early space with this work. And so if that’s you, it’s okay. And I’m really, you know, looking to connect with other folks that are doing this work, especially at sort of more regional organizations.
[Margaux Maxwell] 12:13:00
I would say that we, you know, obviously with the explosion of generative AI and I think with AI coming into the public consciousness in a new way with, you know, obviously some of the marketing that was done around ChatGPT and Anthropic.
[Margaux Maxwell] 12:13:15
Right. We really saw a need to approach this kind of from the newsroom first. And so I really worked closely with our news director and some of our editorial leads. And we also organized.
[Margaux Maxwell] 12:13:31
Uh, this cross-functional AI policy group, um, to shape and continue to shape this work. Um, and we really saw a need to start with, um, the newsroom A to really build that trust, uh, and also to, um.
[Margaux Maxwell] 12:13:47
Really start with a statement of values, which is, I think, where we are, just so we can be clear on that. And once we’re aligned on those values, I think it gives us room to then have those internal and external conversations that we need to have.
[Margaux Maxwell] 12:14:04
down the road. So that’s sort of where we started from, is sort of, you know, our editorial work and our values, and kind of bringing our cross-functionality in as we start shaping that.
[Mike Janssen] 12:14:16
Let’s talk a bit about the processes because y’all took different approaches to — pretty different approaches to how you developed your policies. Drew, Chicago Public Media took a more formal process with an oversight committee, input from many departments. What did that process help you resolve that a simpler process might have?
[Mike Janssen] 12:14:35
missed or overlooked.
[Drew Brody] 12:14:38
It definitely was not a simple process. So, you know, I think we kind of went the opposite of Margaux in some ways, where we started with the cross department AI oversight committee, where we made sure that we had representation from all different parts of the organization.
[Drew Brody] 12:14:53
So that we could…
[Drew Brody] 12:14:55
Look at other policies that existed, uh, so that we could do some research on the technology itself, and that we could start from the place of audience trust, and say, have really broad discussions of where we think it’s important.
[Drew Brody] 12:15:11
what the guardrails are, uh, to be, um… to be transparent with our audience, so that we know, okay, this is… this is… beyond this, we’re gonna have issues with our audience trust.
[Drew Brody] 12:15:24
We worked a draft of the policy from there. We also did a survey of the entire organization to get some feedback and to understand their concerns and some initial feedback. But we created a draft policy from there.
[Drew Brody] 12:15:39
And then we went back to, um, some of our major stakeholders, including, uh, the… our unions. We went to the… the editorial team and the newsroom to then have deeper conversations of, here’s where we think the guardrails are. Now, let’s talk more deeply about your work and how that fits into this overall policy.
[Drew Brody] 12:15:58
and move some of the boundaries based on those concerns. Um, so our first thought was the committee was really for audience trust. We’re then gonna have a separate working group within the newsroom itself in order to.
[Drew Brody] 12:16:13
be a little more granular about the day-to-day. How does this actually affect the work that you’re doing in the newsroom? Um, and then one other thing that I think was important for us is we have split policies. So we have a business side policy that’s really focused on.
[Drew Brody] 12:16:29
information security and privacy, and are these tools trustworthy? Are they protecting our information? Are they using our information to train their models? And so, we have a sort of use case and technology-specific approach on the business side.
[Drew Brody] 12:16:44
Whereas on the editorial side, it really, really was about protecting the humanness of our work, defining exactly which parts of our work need to stay human and completely human, and which parts we think we could benefit from.
[Drew Brody] 12:17:00
uh, some of the AI tools that are out there while protecting audience trust. So that sort of bifurcated approach helped us, um.
[Drew Brody] 12:17:07
helped us…
[Drew Brody] 12:17:09
make the make policies that made sense for each part of the organization.
[Mike Janssen] 12:17:14
Maggie, Little Rock moved faster, and you use a template created by the Poynter Institute for newsroom AI policies. What made that approach workable for the station?
[Maggie Ryan] 12:17:26
Yeah, so like I mentioned, we have a really small newsroom and so we, as all of our reporters, we’re already doing 5 million other tasks throughout the year.
[Maggie Ryan] 12:17:36
the time we were developing this policy, no one really had time to sit down and go through and, like, craft something from scratch. So having a template from Pointer that was already built and kind of gave us a framework to, um, have a conversation and, like, ask the questions of, like.
[Maggie Ryan] 12:17:52
how are other stations or other media organizations using this technology? What conversations might they be having that helped guide our discussions as a newsroom, and one just helped us be a little bit more efficient with our time and.
[Maggie Ryan] 12:18:10
not having to sit down and go through every single, every single AI technology that we could use, but really just see.
[Maggie Ryan] 12:18:20
here’s a few that other other stations are using or other organizations are using. Here’s how they might be using it. What are we comfortable with this newsroom? What are we comfortable with doing? And that allowed us to create a policy, I think, within about a week of starting that conversation.
[Maggie Ryan] 12:18:39
And we immediately had something to put online to put out there so that going forward, if we had any questions about AI in the newsroom or from our audience, we would have something to point to and have something to kind of guide those.
[Maggie Ryan] 12:18:54
future conversations.
[Mike Janssen] 12:18:57
Margaux, OPB’s policy is more concise, I would say, and based more on principles than some of the other policies we’re talking about today. Why did you decide to take that approach?
[Margaux Maxwell] 12:19:10
Yeah, I think, um, so, I mean, first of all, we formed this cross-functional, um, AI policy group, which had representatives from, kind of, different functions around the, um, organization, and, um, I think…
[Margaux Maxwell] 12:19:24
that was a really important space just to be able to discuss kind of what our values are, um, around this technology, um, and I think, um…
[Margaux Maxwell] 12:19:36
you know, I think our feeling was until we kind of have those values hashed out, you know, we can’t really have conversations about anything else. I also think that, you know.
[Margaux Maxwell] 12:19:50
We’re sort of a mid-size sort of to larger organization. There’s a lot to move through. And as we move through the organization and have these conversations, which is sort of the next planned approach that we have, we have a working group.
[Margaux Maxwell] 12:20:05
um, with myself, a couple of editorial leads, and our head of technology to kind of, like, move through the organization, now that we have these values out there, and try to have some of these more specific conversations, and, um, around what we’re doing and what we’re not, and, you know, different… there’s also a range of feelings, you know, out there about AI, and I think.
[Margaux Maxwell] 12:20:25
You know, stating our values really builds a foundation of trust.
[Margaux Maxwell] 12:20:32
As we move through the organization and we have these conversations, you know, things like transparency, things like being people first, right? When we’re talking about something that can be sensitive, you know, my view is that, that trust is really foundational and having those values in front of people just helps facilitate those conversations.
[Mike Janssen] 12:20:50
Yeah, definitely. Kyle, Detroit spent about a year developing its policy. And at the end, you had staff sign the policy. Why did you want that level of involvement and internal review in the process?
[Kyle Vernier] 12:21:06
Yeah, I mean, I think our goal was just to try to get it right. You know, our process is pretty similar to the one that Drew explained, you know, and also Margaux, we formed a small group of people, and I think really our first step was.
[Kyle Vernier] 12:21:22
To get an understanding of AI, you know, none of us were AI experts per se, you know, just people who maybe had interest and so we really spent a lot of time like attending webinars, you know, reading other policies that were out there at the time a lot of organizations were releasing policies.
[Kyle Vernier] 12:21:38
including Nebraska, Connecticut, the New York Times policy had recently come out. So, you know, we read a lot of other policies. Like I said, we attended a lot of webinars, you know, went to summits and things like that, and then also met internally.
[Kyle Vernier] 12:21:55
You know, with departments to see how they were using, how their individual staff members were already using AI, you know, within the company, and then also get their feelings about how it worked. And then we took those things through various, you know, levels of.
[Kyle Vernier] 12:22:12
Of, you know, leadership meetings, a meeting with our leadership team, a meeting with our management team, you know. Discussing it with all of our staff and all with the goal of just making sure that we were, you know, landing on a policy that we all felt like we could stand behind that was.
[Kyle Vernier] 12:22:27
you know, understood, that presented the guide rails that were genuinely useful for our staff. They weren’t just, you know, it wasn’t just a bunch of rules that you would read and then forget. They were things that you could, you know, think about them and understand, like, oh, like, I can use AI this way, or I can’t.
[Kyle Vernier] 12:22:42
And that’s, yeah, we really just wanted to get it right and make sure it was presenting the best policy for our station and for our, you know, our viewership.
[Mike Janssen] 12:22:52
Sure. I want to talk now about some of the specifics in your policies. And Kyle, let’s stay with you. Detroit drew one of the clearest lines against using AI in the policies that we’re talking about to generate or develop editorial content.
[Mike Janssen] 12:23:06
For example, it bars using AI even to generate content ideas. Why was that approach right for Detroit and some of the other provisions that you made about how it can and can’t be used?
[Kyle Vernier] 12:23:18
Yeah, so specifically talking about like the idea generation thing, we, you know, between our community advisory panels, our board, and our staff, we have more than enough ideas, more ideas than we can possibly execute.
[Kyle Vernier] 12:23:33
Um, in a given amount of time. So, um, we don’t really need, like, AI to help us, uh, find what our audience needs. Our audience is telling us what they want to see. Um, and so that’s, you know, really important to us. But then, from the other side, you know, thinking about using it in content.
[Kyle Vernier] 12:23:49
I think you know something that.
[Kyle Vernier] 12:23:54
you know, having a system… I think the thing with AI, when you’re using it in content, is that it requires a lot more scrut… uh, you know, you really have to scrutinize it. It’s an un… uh, you know, it’s an unreliable narrator. And so, I think, you know, any sort of, like, savings that you might be getting on one side.
[Kyle Vernier] 12:24:11
You have to, you know, you’re more than getting it on the other side. And our team is, you know, focused on doing things, you know, according to the PBS standards, you know, journalistic standards. And so we just really found that we didn’t need the tool there.
[Kyle Vernier] 12:24:27
You know, they’re already doing it and it’s adding a layer of, you know, review and approval and, and that, that would just be adding to our process. And so our process works as it is in terms of like developing content and the benefits.
[Kyle Vernier] 12:24:42
that AI might bring to it just don’t outweigh the extra, you know, work that it would need to properly vet all that content.
[Mike Janssen] 12:24:52
Drew, Chicago Public Media landed somewhere different. For example, AI can be used for brainstorming headlines, coming up with synonyms when you’re writing, but it should not be used for drafting or writing content.
[Mike Janssen] 12:25:08
How did you arrive at that distinction and the other provisions about what’s permitted and what isn’t?
[Drew Brody] 12:25:15
Yeah, so first, I want to caveat that by saying that all… any editorial workflow, any use of it, has to be approved by an editor or a manager. So, even if it’s allowed in the company policy, there’s still.
[Mike Janssen] 12:25:26
Hmm.
[Drew Brody] 12:25:29
it would still need to be approved for an individual person. So, uh, and the reason I mention this is because maybe someone who’s newer to the organization or a junior reporter, we might say, actually.
[Drew Brody] 12:25:42
this is actually not right for you to be using the technology in this way. It’s really important for your development and for your work here, for you to be doing this work. But for someone who’s been around for a while, who already has the development, who understands what a…
[Drew Brody] 12:25:57
Good headline might be in what what it should what what it might it.
[Drew Brody] 12:26:01
what… how to use the ideas. I think to Kyle’s point, to… to make sure that we don’t have too much on the other side of… of reviewing the work and making sure that it was, uh, you know, 100% in line with the rest of our standards. We’ve… I think a lot of editors are making the choice that they don’t want to develop new.
[Drew Brody] 12:26:19
Talent that way, but that for people that are more developed, there’s more flexibility. Um, we wanted to create some room for the remarkable and, uh, remarkable technology that AI, uh, brings to the table, especially in terms of.
[Drew Brody] 12:26:35
brainstorming and idea generation. And I sort of hear what you’re saying, Kyle, but I think I personally disagree. I think one of the most amazing things about AI is that it can, uh…
[Drew Brody] 12:26:47
it… it can be someone you have a conversation with, and just having that conversation can help you generate ideas on your own as well. So we wanted to leave room for the… the… that…
[Drew Brody] 12:26:59
those aspects of the technology, while stopping short of writing drafts, or actually writing content. Even though you might say it’s a distinction without a difference to say, like, brainstorming versus writing, but our staff found that there really is a difference, especially.
[Drew Brody] 12:27:16
In… in the mindset here of what it’s… what its core use is.
[Mike Janssen] 12:27:22
Maggie Little Rock also barred some uses. New staff can’t use AI for copy editing, for example, and interns can’t use it at all.
[Mike Janssen] 12:27:32
What risks were you trying to avoid with those guardrails?
[Maggie Ryan] 12:27:36
Yeah, so a few reasons for the copy editing aspect of this. One, as a newsroom, we were really uncomfortable with the idea of putting our work and our reporting into a generative AI system where we may or may not be able to control what the AI does with that information, especially.
[Maggie Ryan] 12:27:53
Starting out with not many people in the newsroom understanding fully how this technology works. And there’s been a lot of concern of how sensitive information and like sensitive stories or sources could be stored or used. So we decided we’d rather err on the side of being really cautious.
[Maggie Ryan] 12:28:11
Um, and we also have in our policy that anything we do, like, put into or pull out of AI must be reviewed by a human copy editor anyway, so we didn’t really see this as a… using AI as a copy editor as a time-saving tool, because on the other end, there would still have to be a human copy editor.
[Maggie Ryan] 12:28:30
Um, and then with the interns, um…
[Maggie Ryan] 12:28:33
a lot of the interns that we’ve had so far have been, have either never worked in a newsroom before or have been pretty fresh to writing in a new, in a, to journalism, to writing for a newsroom. So we wanted to.
[Maggie Ryan] 12:28:49
give them the chance to develop their own skills without using AI as a crutch, or without becoming reliant on it. Also. it’s this huge, complicated thing that’s we’re not even 100%, 100% sure how we’re going to move forward.
[Maggie Ryan] 12:29:05
with it, or what it will become in the future. So, again, we wanted to err on the side of caution and not tackle that beast head-on with interns that are probably only going to be here for a couple months.
[Mike Janssen] 12:29:20
Let’s talk a little about policies where you try to take into account the concerns of staff about AI kind of encroaching on their territory. Margaux, OPB’s policy says AI will not be used to reduce staffing or replace journalists. Why was it important to you to have that stated so clearly?
[Margaux Maxwell] 12:29:39
Yeah, I think because there’s a lot of fear around AI, and I think that stems from a lot of places, but I think one major area where it stems from is AI is going to take my job. And if that fear is dominating the conversation, it really shuts down any room for AI.
[Margaux Maxwell] 12:30:00
innovation or adoption. And so I think both as a signal to our staff, hey, these are not our values around this technology. Your job is not at risk. I think it helps, again, build that trust to be able to have those conversations as we go into these really sensitive specifics that we’re about to enter into.
[Margaux Maxwell] 12:30:23
And so there’s that. And then I also think, you know, it’s important to signal that to the public as well and make sure that they understand that when they support us, they’re not supporting, right, us, you know, cutting a bunch of jobs and, livelihoods and, and replacing those folks with robots. And so.
[Margaux Maxwell] 12:30:41
Um, you know, and I think it also plays into some of the, you know, human accountability pieces, right? That, um, ultimately, you know, uh, we have human beings that are accountable for every way that we use this technology. Um, so I think from a sort of a values, uh, perspective, both internally and externally, and then.
[Margaux Maxwell] 12:30:59
Um, you know, also just, again, that foundation on which we can now have these, um, kind of more weedsy, um, tricky conversations.
[Mike Janssen] 12:31:10
Drew, at Chicago Public Media, how did staff concerns about what work should remain human shape CPM’s policy?
[Drew Brody] 12:31:20
I mean, the good news is, there wasn’t…
[Drew Brody] 12:31:23
There wasn’t a ton of arguing about, in general, that our work.
[Drew Brody] 12:31:29
editorial work should remain human-generated, and that we have no interest in creating work that doesn’t.
[Drew Brody] 12:31:39
uh, originate from a human, so there wasn’t really anyone on the other side of that argument, um, which is… which is, uh, fortunate, I guess. Uh, so then the decisions… the discussions were really around the edges, like, where exactly is that edge, and.
[Drew Brody] 12:31:57
you know, we also said… we also don’t allow any of the editing or the editorial work to happen with AI, and decided that that’s important to stay fully human as well. There… there’s edge cases that we have, like, um, we have some Spanish translation work.
[Drew Brody] 12:32:14
And so we had discussions, like, okay, if we could use this technology in order to increase the amount of Spanish language material that we create by doing first-pass translations, uh, this is work that we would not otherwise be doing, but it…
[Drew Brody] 12:32:28
it meets our mission, uh, should we do that? Um, and so those are the types of discussions that we were having, which are sort of second-level discussions.
[Mike Janssen] 12:32:41
Let’s look a little bit at disclosure and the guidelines around when you disclose use of AI.
[Mike Janssen] 12:32:49
Has your organization disclosed AI used to audience yet? And if so, what does that look like? Kyle, let’s start with you.
[Kyle Vernier] 12:32:57
Yeah, I mean, so, uh, because we don’t use it in content, we, we have not had, um, you know, a public disclosure yet of any sort like that, um, because we haven’t used it to generate any, any content at all. Um…
[Kyle Vernier] 12:33:10
So it’s not something that we, you know, I mean, we have, we talked about it as during our, you know, year long process of developing the policy, you know, but because of where we landed with, you know, our pretty, like you said, hard line about using generative AI and content.
[Kyle Vernier] 12:33:29
We just haven’t felt that we needed to, you know, dive too deep into that part of it yet. So yeah, that’s where we’re at with that.
[Mike Janssen] 12:33:38
Yeah, Maggie, have you guys had to disclose anything?
[Maggie Ryan] 12:33:40
We’re in a similar place as Kyle. We have the policy and we have guidelines for how we would disclose like something like translation or transcription or significant reporting that’s used, that comes out of a publication.
[Maggie Ryan] 12:33:57
using… a reporter using AI for some reason, um, but it’s still… we still only have, like, one or two people in our newsroom that are really experimenting with it, and so far that hasn’t led to any significant, um.
[Maggie Ryan] 12:34:11
That hasn’t led to anything that we felt was needed to disclose to our membership.
[Mike Janssen] 12:34:17
Gotcha. Drew, what about at Chicago?
[Drew Brody] 12:34:22
Uh, yeah, we similarly have our editorial use of AI policy public on our website, and as long as, um, work is done within those cloud finds, we’re not then doing an additional disclosure. Um, we do require disclosure, um, similar to what Maggie said, if.
[Drew Brody] 12:34:39
AI was used in any significant way in the actual reporting, especially if there was anything that would be unverifiable in any way, that was… we’re just taking it from the AI. The only place where we have a disclosure on our website is in the Spanish language content.
[Drew Brody] 12:34:54
if a first-pass translation was done, and it says this was from a translation done by an AI tool and edited by a human.
[Drew Brody] 12:35:04
Um, because it’s…
[Drew Brody] 12:35:07
for those audience trust reasons. Like, we really do want to say if any writing whatsoever is being done with AI, we’re going to disclose it. And that translation, while it’s not original writing, felt like it was on that side where we needed to disclose it.
[Mike Janssen] 12:35:23
And Margaux?
[Margaux Maxwell] 12:35:25
Yeah, again, because of where we are with this kind of early phase of this value statement, we haven’t really — I mean, we’re not currently using generative AI and anything that is public-facing. And we’re not even really using it internally either.
[Margaux Maxwell] 12:35:42
an internal tool as part of Microsoft Copilot, which is kind of already part of our package, just to get folks, you know, familiar and comfortable with AI tools. And so no, because we’re not using it quite yet. We and we’re still figuring out how you know we.
[Margaux Maxwell] 12:35:58
We want that to show up in our work. We are not at the place of disclosure. I will say that we are talking about whether and how we would promote this public AI policy now that we have it. And so I’d definitely love to partner up with other folks that are kind of thinking.
[Margaux Maxwell] 12:36:14
Uh, about that thoughtfully, especially, um, you know, that tandem between, uh, transparency, which we do value and is in our, is in our, uh, statement here, and then community trust and sort of how to facilitate those conversations. So, if you are someone who is thinking about, um, promoting your public-facing AI policy within the community and you want to chat, uh, please, please reach out.
[Mike Janssen] 12:36:36
All right. Well, we’re going to try to end a little early to let people get to that next webinar. As we were discussing, I want to make sure we have time for questions. So I want to get into the chat a bit here.
[Mike Janssen] 12:36:47
uh, throw some of those out. Um, we had one from Emily Colby. Um, do any of you have a set cadence for how frequently you review and update your AI policies, and how are you thinking about what updates might be needed? I think that’s a great question, so let’s, um, let’s start with Maggie on that one.
[Maggie Ryan] 12:37:03
Yeah, we haven’t updated our policy since we first published it back in December of last year. Um, we did just add a couple new staff members in our newsroom recently, and we are expecting to hire another person before the end of the summer, so I and our news director have been talking about when we want to start that.
[Maggie Ryan] 12:37:21
set kind of follow up conversation of bringing the new people on board with this policy, making sure everybody is aware and up to date, and also talking about if anything from the policy that we currently have published online needs to be changed or needs to be added.
[Maggie Ryan] 12:37:36
But right now, because it’s, again, not being used very widely, or either very widely or not at all in our newsroom, we… it hasn’t been a priority for us to kind of go back and check in so far.
[Mike Janssen] 12:37:52
Drew?
[Drew Brody] 12:37:56
Our AI oversight committee meets at least quarterly, but we’ve actually found reason to need to meet more for issues that come up. On the business side, we have a process where people can submit a request to have a tool and use case reviewed.
[Drew Brody] 12:38:10
And then that goes to the IT team, and they review the tool for security and privacy. Um, if they greenlight it, that goes up to the AI Oversight Committee to just make sure it’s in line with our overall policies, and then we add that to our list of approved tools and use cases. So that happens.
[Drew Brody] 12:38:28
it’s not on a regular cadence, but, you know, I would say maybe once a month or once every two months, someone’s submitting a tool that they’re interested in. On the editorial side, we have not updated it since we published it.
[Drew Brody] 12:38:44
the newsroom is currently now going to be starting a working group to actually dig into their day-to-day usage, and they might pop up recommendations to the Oversight Committee to review, but it’s not a… we’ll update it as needed.
[Mike Janssen] 12:38:59
Kyle?
[Kyle Vernier] 12:39:02
Uh, yeah, I mean, so we just, um, implemented a, um, uh.
[Kyle Vernier] 12:39:07
uh, every… at least twice a year, um, gathering of the… of our AI group, um, to review the policy and talk about, you know, tools, and if we feel like the policy needs to be updated. Um, so that’s gonna happen at least twice a year, um, but likely, like Drew said, I think we will be finding needs to meet more frequently than that.
[Kyle Vernier] 12:39:26
Um, to review, um, as things change, you know, the technology’s changing so quickly that, you know, you really do feel like you have to kind of stay on top of it to, uh, to make sure that everything, although use, you know, cases that people might have are, are getting, um.
[Kyle Vernier] 12:39:41
There’s a part of the policy that talks to it, you know? So, yeah.
[Mike Janssen] 12:39:46
Margaux?
[Margaux Maxwell] 12:39:49
Um, yeah, so right now, our forum for that is the AI Policy, uh, group. I imagine it will tighten up a little bit. Right now, it’s pretty broad, especially as we were having this sort of, um, you know, broader, uh, values…
[Margaux Maxwell] 12:40:01
big conversation, and, you know, ultimately, I think we’ll need to identify a smaller group to kind of, um, get that, you know, gather up that feedback from across the organization and continue to update it. Um, you know, and as the technology itself evolves, uh, in addition to our use of it.
[Margaux Maxwell] 12:40:19
You know, like I said, our kind of next steps are the we’re going to move through the organization and then we kind of have, some ideas for, you know, we’re starting with our internal tool, potentially building like a policy bot, an internal tool that helps people sort through.
[Margaux Maxwell] 12:40:35
OPB Policies, you know, something, really low hanging fruit in that area, then potentially a more product development, use case. So one thing we’ve talked about, and it’s not decided, but one potential is, you know, personalization, right? Recommendations are things that people have been getting for a long time from newsrooms. It’s not a shocker for.
[Margaux Maxwell] 12:40:56
audiences to be getting recommendations based on first-party data. We just sort of have a new way of doing it with vectors and embeddings and those things. So potentially some kind of, like, personalization or sort of, like, related, you know, if you enjoyed this, here are some other things you might enjoy. Again, we’re still in the very early stages of exploring whether or not that is something we want to do, and then.
[Margaux Maxwell] 12:41:16
Potentially some kind of newsroom tool around, covering, you know, government meetings. So as we, again, as we sort of move through the organization and we decide how we want to sort of inch closer to a newsroom tool, maybe.
[Margaux Maxwell] 12:41:31
Um, we’re, we’re sort of leaning on that, both the policy group and then the organizational feedback and the mechanisms to sort of work with teams and gather that feedback as we make these decisions.
[Mike Janssen] 12:41:45
All right, I have another question in the chat from Naomi Starobin. Um, are any of you at a university licensee station, and what’s the impact on your AI policy and work? I think of the four of you, the only one that would fall into that camp is Maggie, right, at Little Rock Public Radio?
[Maggie Ryan] 12:42:00
Yeah, so we are licensed to the University of Arkansas Assistant Board of Trustees. That hasn’t really had an impact on either our development or the policy. This came up through our newsroom wanting to implement policy and have something.
[Maggie Ryan] 12:42:16
online for our listeners, for our audience. And we haven’t heard anything from the university leadership about it so far.
[Mike Janssen] 12:42:29
All right. Let me see. We have another good question here from Emily Colby. And I know that it actually — yeah, I’m sorry.
[Chad Davis] 12:42:35
Hey, hey, Mike.
[Chad Davis] 12:42:38
You know, Amber is, Amber’s that university licensee, as am I, but I responded in the chat. Amber, maybe you could speak to that.
[Mike Janssen] 12:42:38
Go ahead.
[Mike Janssen] 12:42:42
Oh.
[Chad Davis] 12:42:45
to some of that question as well.
[Mike Janssen] 12:42:47
Yeah, go ahead.
[Amber Samdahl] 12:42:48
Oh, sure. Sorry, sorry, Mike. Yeah, I think it’s a really good question. I’ll say, so we’re part of UW Madison and one of the things that I’ve noticed is different in how we’ve approached our AI policies. UW has very strict like cybersecurity rules.
[Mike Janssen] 12:42:50
No worries.
[Amber Samdahl] 12:43:04
um, that cover AI tools. They have a list of approved, um, like, already integrated into our software tools, um, AI tools, like we’re all in Gemini, we have access to, you know, ZoomAI, Companion, things like that.
[Amber Samdahl] 12:43:19
So we were actually able to build our policy in Wisconsin on kind of the foundation that the UW already had, which actually helped us in the, I would say, in the cybersecurity space in particular, because they already had that there for us to build upon. So we really focused on what uniquely public media was.
[Amber Samdahl] 12:43:36
Bringing in addition to all those base AI rules.
[Mike Janssen] 12:43:42
All right. Thank you. We have a question from Komaltra Vedi. I’m interested to learn more about the vetting process you’ve been using for the AI technologies you implement use of. How did you define that? And was there a framework that you use in order to feel that you have securely vetted the technology?
[Mike Janssen] 12:43:57
Anyone able to take that one?
[Drew Brody] 12:44:02
I could say that our IT teams do this, and they, um…
[Drew Brody] 12:44:09
Basically, they’re checking to see, is our information being used to train models? Do they explicitly say that they’re not used to train models? And so anything that is, we don’t allow, at least not for anything that’s work-related.
[Drew Brody] 12:44:26
Um, or anything, you know, proprietary whatsoever. Uh, and then…
[Drew Brody] 12:44:32
Yeah, and then we have certain guardrails around, even for things that are not used to train, about, um…
[Drew Brody] 12:44:40
aspects of their privacy policies and other pieces there. But mostly, it’s about the security of the data that we’re uploading.
[Mike Janssen] 12:44:48
Anyone else?
[Mike Janssen] 12:44:53
OK. This one I know applies to Oregon Public Broadcasting. This is another question from Emily Colby. Did any of your newsrooms consider AI’s environmental impact, and how did that factor into developing your policies? That’s actually explicitly mentioned in OPB’s policy. So Margaux, do you want to take that?
[Margaux Maxwell] 12:45:10
Yeah, so the language we have is we acknowledge the environmental impact associated with the use of AI systems and commit to being mindful of resource efficiency in their application. And again, because this is a foundation on which we’re going to be having both internal and external conversations.
[Margaux Maxwell] 12:45:28
entirely worked out what that looks like. I think, again, it’s about the commitment that it will be part of those conversations as we make those decisions. Something that I can foresee that is not on our specific roadmap at OPP at all yet is.
[Margaux Maxwell] 12:45:45
Or going back to the personalization example, like if we were to build some tool with that, and we were to go through ChatGPT’s API, for example, considering that tool over another.
[Margaux Maxwell] 12:46:02
Um, in order to build that out, um, using that first-party data.
[Margaux Maxwell] 12:46:07
Our commitment here is really that, you know, those considerations would be part of the conversation as we make those decisions.
[Mike Janssen] 12:46:16
I don’t remember that anyone else’s policy mentioned that explicitly as yours did, but did anyone else consider this behind the scenes that you’d want to chime in on this?
[Drew Brody] 12:46:26
I see…
[Kyle Vernier] 12:46:26
Yeah, PBS, we have an environmental news program, Great Lakes Now, and they cover data centers. And so, you know, this was part of the conversation that we had.
[Kyle Vernier] 12:46:36
But because like we don’t allow, you know, in content, we don’t allow generative AI for the creation of video or audio or images, some of the really heavy uses, we didn’t feel like it needed like to be explicitly spelled out in the policy.
[Kyle Vernier] 12:46:52
But these are conversations that we had with the people on that team and our content team in general.
[Mike Janssen] 12:47:01
Drew, did you want to say something?
[Drew Brody] 12:47:02
Yeah, I’ll just… I can say for sure that this is on people’s minds. This was a… from a values perspective, very important to a lot of people in the organization.
[Drew Brody] 12:47:13
we had a lot of conversations about this, um, and we studied a bit what our actual… what our actual contribution would be from the uses that we were contemplating, and it’s, like, negligible from an individual organization standpoint. Um, which…
[Drew Brody] 12:47:29
so… so it’s less about what we would be contributing to the environmental impact, and then more about, uh, the values of the impact of the industry as a whole. Um… we don’t make a ton of other decisions that way, you know, whether to…
[Drew Brody] 12:47:43
serve meat at company lunches and things like that, you know, so… so we just, um, decided that our role is to report on these things, to investigate and to make sure that we are covering the overall industry environmental impact that that, um.
[Drew Brody] 12:47:59
we weren’t going to, uh… and we, again, we would be mindful about the usages, for example, where we could set up local, uh… local tools that could potentially save our usage there. Uh, we would do that, but otherwise, we were…
[Drew Brody] 12:48:14
We were thinking that our role is really to be the investigators in the reporting on the overall impact there.
[Chad Davis] 12:48:25
Mike, I’ll just jump in here real quick. It’s something that we actually wrote into our new policy as an appendix, and I’ll just stick that.
[Chad Davis] 12:48:36
People can get the link there for that. That’s Appendix B. Where we’re kind of coming down on that is to encourage people to start thinking about AI.
[Chad Davis] 12:48:48
tools the way they would a ream of paper, or a gallon of gas, to kind of begin to apply more of a scarcity mindset to it, because it’s been pitched to everybody for 3 years as, like, here is this limitless technology, and.
[Chad Davis] 12:49:04
And people are just doing what they call slot machine prompts, where they’re just throwing things in, they don’t like it, just hit the button again, just hit the button again each time you hit…
[Chad Davis] 12:49:14
one of those buttons, it has an environmental impact. And so we want to change the thinking around it so that people are conscious. And it kind of builds a little bit off what Margaux was saying of, you know, use the right tool for the job.
[Chad Davis] 12:49:30
like, you don’t have to use the big frontier models for every single thing you do. Um, someone jokingly said on Twitter at some point, you know, you don’t need God to write your emails for you. You know, there are better tools, so try to…
[Chad Davis] 12:49:46
Try to use the one that is appropriate to the task that you’re trying to do. And then be aware, if you’re gonna generate video, that’s gonna have a much bigger environmental impact than, you know, generating text or something along those lines.
[Chad Davis] 12:50:03
Those those lines.
[Mike Janssen] 12:50:05
We had a question from Tim Olson regarding images such as creation or augmentation. What are some of the most thoughtful policy language folks have seen? Kyle, why don’t you take that since you’re at a TV station?
[Kyle Vernier] 12:50:19
Yeah, I mean, this is complicated ’cause we don’t allow that. And so I would say that like, you know.
[Kyle Vernier] 12:50:25
A really good resource for this is the PBS Editorial Standards. So, you know, we had conversations with Talia Rosen on the PBS Standards and Practices team about this. You know, the standards themselves don’t.
[Kyle Vernier] 12:50:42
prohibit, you know, the creation of, you know, images or sound, but there’s so much concern about, you know.
[Kyle Vernier] 12:50:51
you know, the way that you’re representing reality, or that you may be representing or augmenting reality, that it makes it very difficult to, like, you just have to be very, very conscious of the way that you’re using those tools.
[Kyle Vernier] 12:51:07
to make that kind of content. And like I said, for us, it just was, you know, too much, you know, to monitor all of that. But I think that there are, you know, PBS, you know, they obviously, you know.
[Kyle Vernier] 12:51:24
get documentaries that use that content, they’re reviewing it, they’re having conversations with the producers, and they include conversations like, what is the prompt you use to generate this image? Or, what is the prompt that you use to alter this image? You know, so asking the producers to maintain those records so that you can, you know, vet that.
[Kyle Vernier] 12:51:43
That you’re, you know, that it’s something that you think is still representative of the truth. And so I think like that, you know, those are, it’s a good resource. The PBS standards are a good resource for looking at how to properly do that.
[Kyle Vernier] 12:51:58
But I think it is, it’s a lot. I think it’s a lot to take on, and that’s why we decided, you know, not to.
[Mike Janssen] 12:52:06
Has anyone else considered this in particular?
[Drew Brody] 12:52:06
We also…
[Drew Brody] 12:52:09
We also do not allow almost any generative AI in image, in visual content. The only thing that we do allow is if there’s some sort of illustration and they need to create assets.
[Drew Brody] 12:52:25
for that illustration. So, you know, generate, like, a background, or a shape, or something like that. Um, but, uh, that’s it. Nothing that could be mistaken for, uh, original photography, nothing that is representing a real person.
[Drew Brody] 12:52:43
place, building, location, anything that could be taken as real is not allowed by our policy.
[Mike Janssen] 12:52:53
Anyone else?
[Chad Davis] 12:52:54
Drew, can I just ask a follow up? So is the line you’re drawing, ’cause we’ve talked a little bit about this here at our station, is the line you’re drawing kind of.
[Chad Davis] 12:53:01
faking or imitating reality, like, if it is animation, if it is…
[Drew Brody] 12:53:04
Correct.
[Drew Brody] 12:53:06
It’s both…
[Chad Davis] 12:53:06
Yeah, so if it’s not trying to…
[Drew Brody] 12:53:09
It’s that, and also, even in non-realistic things, we can’t generate, like, a completed.
[Drew Brody] 12:53:18
abstract image, either. Like, only assets in that image. So even… even for something that’s not reality. But that… that’s the hard line, is, like, anything that could be…
[Drew Brody] 12:53:28
representing reality is not allowed by the policy.
[Mike Janssen] 12:53:36
Just one last question. Emily Colby asks, what do your policies say about on-air disclosure? Does anyone have any kind of — I don’t remember discussing that with anybody, but any provision that you will discuss this on your airwaves or on broadcast?
[Drew Brody] 12:53:56
We haven’t had a need yet, so we would if there was ever any involvement in that, but that has not come up for us.
[Mike Janssen] 12:53:57
Okay.
[Mike Janssen] 12:53:58
Yeah.
[Mike Janssen] 12:54:04
Gotcha. But they are… you all have posted them on your website, right? Or you’re going to, I believe?
[Drew Brody] 12:54:11
Yes, yes.
[Mike Janssen] 12:54:11
OK. Great. Well, we are getting very close to the top of the hour. And I know that we want to try to —.
[Mike Janssen] 12:54:18
Bring things to a close a little early.
[Chad Davis] 12:54:19
We do. My…
[Chad Davis] 12:54:20
Could we grab Rema’s question real quick? She has a good one, as she often does. She specifically talks about the fact that she uses AI to help with writing because she’s dyslexic. So AI is an accessibility tool, and so that helps with grammar, spelling, sentence flow.
[Mike Janssen] 12:54:24
Slate, go ahead.
[Chad Davis] 12:54:38
So, which all of that makes sense to me. Her question is, how is this handled in AI policies? I know with ours.
[Chad Davis] 12:54:48
We don’t allow for inventing of quotes or things like that, but it, you know, we do allow the use of AI for writing.
[Chad Davis] 12:54:59
In many instances, and editing, it’s just not there for, like, you know, kind of changing your words after the fact, that sort of thing. I don’t know, Amber, if you guys are in a similar space, but…
[Amber Samdahl] 12:55:15
Yes, very similar. I’d be curious to hear if others have explored that.
[Drew Brody] 12:55:23
Um, I would say for grammar and spelling, there are tons of tools that people have been using for a long time, even built into, you know, Word or Google Docs that would correct spelling or grammar that I don’t think would be considered generative AI. I think I’d have to know more about what you mean by sentence flow.
[Drew Brody] 12:55:41
I’m not sure what that would encompass, but that would potentially be on the wrong side of our policy, but for accessibility reasons, maybe there would be some discussions about it, but, um, the sentence flow part makes me nervous.
[Chad Davis] 12:56:00
I think certainly your HR departments, just to say this to the, like everybody who’s tuned in right now, your HR departments probably would have thoughts or comments on accommodations.
[Drew Brody] 12:56:11
Yeah.
[Chad Davis] 12:56:12
and things like that. So to me, like, that makes sense that, hey, maybe they’re an exception.
[Chad Davis] 12:56:19
options for these sort of things.
[Amber Samdahl] 12:56:22
I would, I would just add for us, we talk a lot about in.
[Chad Davis] 12:56:24
Yes, please, I’m fine.
[Amber Samdahl] 12:56:25
Internal disclosure, too, like, you know, sharing with others the tools you’re using, um…
[Amber Samdahl] 12:56:31
Just helps open up that transparency even internally is important.
[Drew Brody] 12:56:33
Yeah, thank you.
[Drew Brody] 12:56:35
one thing I’ve seen, just to that point, on the business side, where there’s, um, a lot more experimentation happening here, is people have gotten a lot more comfortable saying.
[Drew Brody] 12:56:46
hey, I took the notes from our meeting, and I, um, put it into our enterprise ChatGPT, which is allowed by our company to, um, sort of…
[Drew Brody] 12:56:57
summarize, like, or to help me put an FAQ together, or something like that. And I think it’s been helpful, and it’s helping people in the organization see what’s possible and ways to, um, be more efficient in some of the work they’re doing, especially some of the, I would say, like, lower.
[Drew Brody] 12:57:13
lower-value, higher-time consumption tasks. And there’s a lot of openness about that, and it’s nice to see people saying, here’s how I did this thing, review it, because then it tells me also, okay, I need to read this as if.
[Drew Brody] 12:57:28
this came from, uh, from AI, and I need to look at that a little bit differently, and assume that other person’s been, you know, then assume that it came from that person. So, um, I do think, as a rule, um.
[Drew Brody] 12:57:40
Even for things that are allowed in a policy, it’s great to be open and transparent about how you’re doing things for the people that you’re working with.
[Mike Janssen] 12:57:53
All right. Well, I think we’re going to be wrapping it up. And I just want to say thanks to everybody. This has been a great discussion. There’s lots more we can get into, but I hope this has been helpful to everybody. And check out those policies we shared if you would like some examples of what these stations are doing.
[Mike Janssen] 12:58:07
Chad?
[Chad Davis] 12:58:09
Amber, you want to bring us home?
[Amber Samdahl] 12:58:10
Sure. Thank you. First, thank you to all the panelists for your time. Really appreciate hearing all of your different perspectives and different approaches. Before we wrap, I do want to again put a plug for that resources page. Not only will you find all of the policies that were shared here today.
[Amber Samdahl] 12:58:26
There’s also an invitation for you all as attendees to share your policies, and we’ll get them posted on there as well. We’re also in the process of collecting AI use case examples from across public media system. We’re just at the early beginnings of that sharing what we have collected through our webinars.
[Amber Samdahl] 12:58:44
So we have about 50 odd examples in there. But again, there’s a form on the page or reach out. We’d love to collect and share all these examples across the system. So I invite you all to check that out and also put in a link for our website.
[Amber Samdahl] 12:59:00
our August webinar as well.
[Mike Janssen] 12:59:02
Yeah, and Current is always looking for people to share their examples of how they’re using AI. So if you have ideas, just send us an email or contact us through our website, and we’d love to hear what you’re up to.
[Amber Samdahl] 12:59:14
Sounds great. Thank you, everybody. Have a great afternoon.
[Mike Janssen] 12:59:15
Thank you.
[Chad Davis] 12:59:17
Go raid the PBS webinar. We’ll see you next month.
[Drew Brody] 12:59:21
Thanks, all.
[Chad Davis] 12:59:31
Thanks.
