Community IT Innovators Nonprofit Technology Topics

Nonprofit AI: Google Chrome Skills and the Center for AI Safety

• Community IT Innovators • Season 7 • Episode 32

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 21:46

This week, Carolyn Woodard covers two resources: a practical new browser feature for Google Workspace shops and a nonprofit watchdog organization working to make the AI industry more accountable. Whether you're trying to get more out of your existing tools or looking for credible resources to bring to a board conversation about AI risk, this episode has something useful.

This episode covers:

  • Google quietly rolled out a feature called Skills inside Chrome — it lets you save a prompt once and run it on any web page with a click, without retyping. Pre-built Skills are also available for common tasks like summarizing long documents or comparing information across open tabs, no prompt-writing required.
  • Skills is a reading and analysis assistant, not an agent — it won't take actions like making purchases or browsing on your behalf. It reads what's already in front of you in the browser and helps you process it faster.
  • If your nonprofit is in the Google ecosystem and staff are hesitant to write prompts from scratch, Skills' pre-built library is a low-barrier starting point. Useful use cases include reviewing foundation grant pages and comparing information across multiple sites.
  • The Center for AI Safety (CAIS) is a San Francisco-based research and advocacy nonprofit whose mission is to reduce societal-scale risks from AI. Their website (safe.ai) isn't a vendor — it's an independent watchdog with accessible explainers, free courses, and fellowship programs.
  • CAIS offers a free AI Safety, Ethics and Society course that's relevant for nonprofits building AI literacy on a budget, plus a fellowship for people doing advocacy work around AI governance, bias, or data center impacts.
  • When staff or board members are skeptical about AI — or when you need a credible outside voice for your AI strategy conversations — CAIS is a more trustworthy resource than asking the AI companies themselves what's safe.

Resources Mentioned:

_______________________________
Start a conversation :)

Thanks for listening. 


Carolyn Woodard

Hello and welcome to the Community IT Innovators, nonprofit AI podcast, midweek check-in. My name is Carolyn Woodard. I am the outreach director for Community IT. I run the webinar program. I host the podcasts.

Carolyn Woodard

I am not an AI expert. Right now, nobody really is. And even if you are an expert in one area, you can't be an expert in all of the areas. So the purpose of this midweek check-in is just to share some stories and resources and news that could be relevant for nonprofits wherever you are in your journey. If you just starting out want to know what AI is, if you've been using AI for a while and want to know how to use it better, we're happy that you're here and thank you for joining us. So

Carolyn Woodard

Today I have just a couple of stories that I noticed that I wanted to share. So if you are using one of the big AI platforms that nonprofits are using, so Copilot in Microsoft, Gemini in Google Workspace, or Claude, which is a standalone and can work with either of them. In fact, all of them can work with any of them, basically.

Carolyn Woodard

You know that those platforms are moving fast. They are releasing new updates, they're releasing new features, they change the menu around. So you're like, wait, I remember this was a tab over here on the left, and now it's a tab on the right. So I would like to try and keep you updated with that. And uh, you know, it's hard to track all of it yourself. And exactly as they're adding new features, it can be hard to know what that little click will do. Uh,

Carolyn Woodard

Google just made a change to their Chrome browser that I wanted to mention. Uh, they didn't do a big fanfare, but I thought it was really interesting. And then I also want to talk about an organization that is asking a lot of big questions behind AI. Who's keeping AI safe and what does that mean for nonprofits? So

Carolyn Woodard

Google rolled out a new feature called Skills inside of their Chrome browser in April. It is built on top of the Gemini integration that's already in the browser. Now, this is something I have to admit, I don't usually use AI in the browser very often. I tend to gravitate toward the standalone apps, desktop apps. Um,

Carolyn Woodard

I have used uh Copilot within Microsoft Tools. So if you open Word, there's a little, you know, the old star pattern, and you can click on that, and it Copilot will ask how it can help you with whatever you're trying to draft. Or if you're in Excel, same thing. There's a little icon and you can get uh help within the tool. Uh

Carolyn Woodard

If you're using it with work, which we always recommend, you should be logging in with your work email and you should have a paid version if you can. Uh, and those paid versions can be lot, there's lots of different tiers. Um, so you don't have to invest in the enterprise version for everyone at your office, but you might want to get just a personal version for yourself. Always check and see if there's a nonprofit discount, of course.

Carolyn Woodard

Apparently, there's a whole world of people who use the AI within their browser. Uh, so either the Gemini integration that's built into Chrome or the Copilot integration that's built into Edge. And so this new feature that Google's rolling out is uh called Skills. So the idea is that instead of retyping an AI prompt every time you want to do a recurring task in your browser, you can save it as a Skill and you can run it on any web page with a click.

Carolyn Woodard

Google has also created some pre-programmed Skills. So things that you might find yourself doing. And those you don't even need to create it. It you just click on it and it will do it in the web page that you're looking at.

Carolyn Woodard

So the way it works, as I said, you write the prompt once, you save it as a Skill, and then you can trigger it. There's a little plus button that you can use while you're browsing. It can work across multiple tabs that you have open at once, so not just the page that you're on.

Carolyn Woodard

So if you're not comfortable or don't totally understand how to save a Skill that you've prompted, you don't have to start from scratch. You can use their pre-built ones if it's something that you're looking for. And I'm sure they're going to keep adding to the library as well. So

Carolyn Woodard

This is different from opening Gemini in a Google window and using it, or using Claude in a separate window in your Chrome browser. So

Carolyn Woodard

When you use an AI tool in your desktop app, or you open a tab and open Gemini and use it through the browser, you're working kind of beside the web. You are inputting, you have to cut and paste the input, or you have to input your question to the fields to have it answer your question.

Carolyn Woodard

It goes out as we've talked about, it analyzes the packets, goes lickety split to the data center, figures out which packets of words come back to answer that kind of a question and what type of information you're looking for based on the other times that that question has been asked and answered. And, you know, using its built-in learning and model, it is returning you this answer.

Carolyn Woodard

The browser native AI is built into the browser. So it can see what you're looking at in real time. It can see the website that you're looking at, quote unquote, see. It's not a person, it's not looking at it, but it's able to analyze what's on the website that you're looking at.

Carolyn Woodard

Skills is Google's formalized version of what Copilot has been doing in Microsoft Edge for a while. So if your nonprofit staff are mostly in Chrome, you're a Google Workplace shop, and a lot of people are comfortable using Chrome, you're on devices that are using Chrome, uh, not Edge, then this is an upgrade. So it's trying to reach what Edge has been doing. And it may even be uh leapfrogging a little bit. It's trying to um to be fine, parity, be as useful. So

Carolyn Woodard

Some examples from Google's library are that you can summarize lengthy documents or lengthy web pages. You can compare product specs. So if you're looking at a new TV on different sites or different TVs, it you can have all those tabs open and then uh click on this Skill to compare them and it will look at those pages, tabs that you have open and give you a comparison. Uh, it can calculate nutrition information from a recipe page. So there's that, that's the sort of thing that these Skills can do.

Carolyn Woodard

It is a reading and analysis assistant, so it's not an agent. I was afraid when I first heard about this. Is it going to shop for me? Like, is it going to find credit card information, you know, and and uh it has my address, and I've been using the Chrome browser to shop, and it's just going to, you know, I'm going to look for black pants and it's going to find them, order them, and they're going to show up at my door. So that is not what this is.

Carolyn Woodard

There are agents that can do that, but that is not what Skills can do. You are still doing the browsing, you are doing the decision making, it can make it faster for you to process what's in front of you or analyze these sorts of things instead of having to look through the different, especially when you're making a decision about a purchase.

Carolyn Woodard

There is an agentic version where the Gemini agent can take actions for you, like book a reservation or make a purchase, but that is a separate feature. It's not being rolled out to everyone, which you know, thankfully, you have to have that be part of something that you've purchased. And I'm sure they make it very clear that it's it will be shopping for you. Um, so

Carolyn Woodard

I find all this interesting because about it's probably nine months ago, so the end of the summer last year, I was going to meet a friend on a weekend on a Sunday. Um, but my friend could only meet me at nine o'clock in the morning, and we wanted to have brunch. So, as a person, a human, I know that a lot of places aren't open yet. They open at 11 on Sundays. So

Carolyn Woodard

I had the bright idea to ask Gemini because I figured Google Maps is right there. So I asked Gemini, here's the metro station where I'm meeting my friend. Uh, can you give me a list of five restaurants that are within walking distance of this metro station and open at 9 a.m. on Sunday? And Gemini very confidently gave me a list of five different restaurants.

Carolyn Woodard

And as I clicked through just to check, none of them were open at 9 a.m. They were all walkable, they were all close to this metro um station, but none of them were open. Just a quick look at the Google Maps listing for those restaurants that had the hours could show that they weren't open. So I was very deep disappointed. I thought, you know, Gemini just isn't up to this task. Um, and I, you know, I didn't really, it's

Carolyn Woodard

It's a natural task to ask that I expected an AI to be good at doing. So I don't want to look at all of these hours that are available in Google Maps. I want the AI to look at those hours and give me only the restaurants that are open at nine.

Carolyn Woodard

But especially at that time, nine months ago, Gemini and the other AI tools, it wasn't actually checking a live website or using the Google Maps data. It was just pattern matching. It was guessing. It could find restaurants that had been referenced as being close to that metro station, but it was guessing at when they would be open. And it had no idea that Sunday at nine o'clock is a problem, which was the whole point of why I asked it that question.

Carolyn Woodard

So this is a bit of the gap that the native AI in the browser is trying to be able to close. It can see what you're seeing on the web page in front of you, and it's reading that content instead of guessing.

Carolyn Woodard

So it's not perfect. And I didn't really expect it to be perfect when I originally asked this question because you know, if you've looked at Google Maps, like businesses don't update their hours. Uh, restaurants particularly, they might be closed for a holiday, they may be closed for a family emergency, and no one updates the hours because it's not what you do. You know, if you're going to the restaurant in real time, you just look for a different restaurant if it's closed and you didn't expect it to be. So, you know, AI can't help with that.

Carolyn Woodard

If the website isn't right, then AI isn't going to, even if it's reading it, it's not going to be able to give you the real information. But being able to read what's actually on the website is an improvement over the hallucination that I got of like, here's five restaurants, you know, any of them will work, and none of them would.

Carolyn Woodard

I will say though that as far as I understand it, the AI Skill still is not able to browse for you. So you can't ask it to go out and and give you return websites to you that will have this information. You would be finding those tabs and then asking it to do the analysis across the tabs of the websites that it can quote unquote see. So

Carolyn Woodard

It's getting a lot better at analyzing content. It still is going to have problems finding accurate information and you know, confidence. The AIs are tools are very, very confident. They're still very confident. Even when you ask them to tell you when they think they have a 95% chance of being correct on something, they'll seize it.

Carolyn Woodard

So if you have staff, if you are in Google Workspace and you're using Chrome, as I said, you know, this is a really good feature to explore. Just see if it's something that can help you. If you're doing, you know, analyzing something that's a long document, like a privacy policy or terms and conditions, it can help you with what it can actually see in the browser.

Carolyn Woodard

If you have staff who are not confident in being able to write a prompt or save a skill, the fact that they have this library with these very common things that you might want it to do, you know, you can just look through that and see if any of those would be useful in something that you're doing.

Carolyn Woodard

Maybe you are doing donor research or you're looking at um foundation information on grants. So being able to compare across websites use cases where this built-in AI to your browser could be helpful to you as a nonprofit, especially if you're already in the Google ecosystem. Um, but

Carolyn Woodard

When AI tools fail, it is interesting to understand why, uh, which is something you can ask an AI tool. Like, what like uh when I had the restaurant fail, I kind of chided it. I was like, well, none of these are open. And of course, being uh AI, it said, Oh, you're right. Thanks for pointing that out.

Carolyn Woodard

But you can ask it, you know, why did you return those results? And uh how did you do the analysis? Uh, so it can tell you why it made something up or made an assumption or guessed. Understanding those limitations can also help you as you go forward know when to trust, you know, what it's returning to you, how often you need to verify stuff, especially if it's something new that you're asking it to do. It hasn't done it before, then you need to.

Carolyn Woodard

I heard someone say treat it like a an intern, a very brand new intern, and you have to check everything that they do. So I will include some of the stories about this. Uh, there is a good how-to-use Skills in Google Chrome uh on Wired. Uh so if you have a subscription there, uh you can uh have a look at it. Also, Google, of course, has information on their blog about it. Uh,

Carolyn Woodard

The second resource I wanted to share is a website, an organization called the Center for AI Safety or C-A-I-S pronounced CACE. The website is safe.ai. It's based in San Francisco. It's a research and advocacy nonprofit, and their mission is to reduce the societal scale risks from AI through technical research, field building, and policy advocacy. So

Carolyn Woodard

They are behind a uh statement on AI risk that came out a couple of years ago that a lot of researchers and executives signed saying that the AI risk was in the same realm as pandemics or nuclear weapons. So you may have been seeing stories about uh people in the AI industry who have become convinced that the AI will become sentient, that it is going to enable really bad impulses among uh humans and potentially do really bad things to the planet, and not just through us creating the data centers, but through the AI itself. So if you're in those doomsday uh scenarios, um, I try not to read too many of them because I find them really uh frightening. But uh

Carolyn Woodard

This website, I think, is uh really interesting because it's not a tool, it doesn't have its own tool, it's not a vendor, it's a nonprofit, it's a watchdog. And they have a lot of resources on their site that you might be interested in looking into.

Carolyn Woodard

They have some really accessible explainers on that AI risk written for a general audience. Um, so you know, if you're worried about those doomsday scenarios, you can have, or if you have people at staff at your organization who are really worried about AI and if they use it, it's going to somehow enable it to do these bad things. Um you might want to have a look at these resources that explain the types of risks they're looking at.

Carolyn Woodard

They offer a free AI safety, ethics and society course that you can sign up for right on their website, which is really relevant for nonprofits that want to build AI literacy, and it's free, so that's a good price point.

Carolyn Woodard

They have a fellowship program, the AI and Society Fellowship, for people researching the societal aspects of AI. So if you do advocacy work, um if you're working on advocacy around data centers, around bias in AI, around the other uh issues that we're running into, uh, if you're interested, as a lot of us are, in insisting that the tech companies build safer AI, that it's safer for users, that it's less exploitive, that it might not have the nudified tools, for example, that they're using in elementary schools. Uh, insisting on those types of regulations and those types of tools that would be a lot more secure and a lot have a lot more safety built into them, then this uh fellowship might be interesting to your organization as well. So we can look into that. But um,

Carolyn Woodard

I just wanted to talk about it because you we talk a lot about the tools and about AI literacy and about exploring and trying. And uh we're talking about these tools built by these huge companies. So CAIS is an organization, a watchdog, and there are other organizations like it, especially in the EU, that are actively trying to influence the companies and the government regulations around them. So I just think it's a good resource. I loved poking around on their website today, so I uh would recommend taking a look at that.

Carolyn Woodard

It makes you a smarter consumer. Your nonprofit can have a more credible voice when you're talking about these AI aspects that you want to be careful of around the communities you care about. Or on the other side, if you have board members, executives, staff members who are concerned about using AI at all, this website has a lot of good ways to talk about what the risks are and what the issues are. So if you can point a board member to that, um, it's not asking Claude or asking Gemini or looking at what Google says about how safe it is or what Microsoft says about how safe it is. Uh, it's an external um watchdog, so it might be a better, more trustworthy resource.

Carolyn Woodard

I will share those links also with you about uh the risk overview and the free course that they have in the show notes. And I just wanted to, you know, share a couple of resources with you. Um,

Carolyn Woodard

Google Chrome Skills is a really practical feature. So if you're using Google and you're using AI in your browser or interested in seeing what AI in your browser can do, have a look at that. Um, the

Carolyn Woodard

CAIS, the Center for AI Safety, uh is a good resource for the bigger picture of where this industry is going. And if you're having those types of big picture conversation with your board, with your funders, with your executives about your AI strategy and your governance, uh that might be a good resource for you as well.

Carolyn Woodard

So as always, AI isn't slowing down and it isn't going anywhere. It's our job to stay curious, stay informed, and keep asking and demanding together, collectively, what we want these tools to be like as we move forward. So I want to thank you so much for joining me today. I will be back in your podcast feed on Friday with our regular technology topics topic, and back here again on every Tuesday with some more uh news and resources that you can use around nonprofits and AI. So until then, take care.