Irene Liu: Welcome to AI Sidebar. I’m Irene Liu, your host and executive director of the AI Initiative at Stanford Law School. Today, I’m delighted to welcome Heather Meeker, an internationally recognized expert in open source software licensing. Heather is a partner at TechLaw Partners and co-founder of Chinstrap Community.

She is also the author of Open Source for Business, a leading resource for leaders navigating open source licensing. In her recent TED Talk on why we need open source AI, Heather explored why openness, collaboration, transparency, and public trust are so important to the development of AI. And that conversation feels even more urgent today as open weight models gain momentum and Chinese models reshape the competitive landscape, and companies and policymakers wrestle with what open source AI actually means.

So today, we’ll discuss why open source AI is having its moment, the difference between truly open source and open weights, whether open models are safer than proprietary systems, and how legal leaders should evaluate the regulatory security and geopolitical questions surrounding these technologies. So, with that, let’s jump into the conversation.

Heather, thank you so much for joining us here at Stanford Law today. Heather, you’ve spent decades working on open source, but the terms become increasingly complicated in this age of AI. So, let’s start with the basics. What does open source AI actually really mean? 

Heather Meeker: Well, if you expected a short answer to this question, I’m sorry, you’re gonna be disappointed because it’s actually a very complicated question.

If you go back to what open source was created to describe, that was software, and as we all know, it worked astonishingly well for software, right? The whole world runs on open source software, and it’s really great in many ways. But translating that concept to other things beside software is actually quite challenging.

[00:02:07] When we talk about AI, first of all, what we’re usually talking about is machine learning models. And the problem with applying a term like open source AI to a machine learning model is that there’s no source. 

Irene Liu: Yeah. 

Heather Meeker: In other words, open source referred to source code for software. So, you know, you had binaries, like executable files, like on your phone or computer or whatever, and those, your computer knows how to run, but those are written by humans in programming languages. That is not how AI works at all. 

AI is trained, and as a person I know charmingly put it, they are big piles of linear algebra. They’re not binary translations of source code at all. Now, you have to use a lot of software in order to use AI, so there, there is software lurking in there. But when people say open source AI, what they’re usually referring to is applying the concept of open source to machine learning. Now, that is very difficult. 

Open source generally was created in order to have software be reproducible and transparent. So, in other words, you could study it and see what it was doing, and that was a really great benefit. And then also, if you wanted to tweak it, you could tweak it, and you could change it, and then you could recreate the thing if you had been given the right source code out of artifacts to do that.

Well, neither of those assumptions works for AI, because there really isn’t a such thing as transparency into machine learning models. At this point, they’re basically a black box once they’re created. 

And also, it is extremely difficult to reproduce them, even if you have the original materials that were used to create them. There are people who have tried to come up with definitions of open source AI, most notably the Open Source Initiative, and what they focused on was the software used to train the model, the data used to train the model, and then the weights or the machine learning model itself. 

And so, one idea of open source AI is you have to say what software you use to train it. And by the way, that’s the easy part because almost all AI is made with open source software that’s freely available. 

Then you have the weights, and the weights are non-transparent by definition, but you can grant people rights to use them in different ways. And then you have the data, and the data is really the sticking point because for the kinds of models that we care about these days, they are created using massive amounts of data.

And so the bid ask on most people’s definition of open source AI is, do you have to provide the data or do you have to just describe the data? And providing all the data is probably functionally impossible for most of the models that anybody cares about using. So that is one of the real sticking points.

So, when people say open source AI, they can mean a lot of things. Sometimes they don’t really know what they mean, but that is basically what it means. Is the training software available? Do you have rights in the model weight, and what rights or information do you have about the data? 

Irene Liu: So, one of the key things that you said earlier that’s really interesting is that with open source AI, it should technically have open source. The source should be available, the source code. 

Heather Meeker: Yes. 

Irene Liu: But source is not available. The source code is unavailable- 

Heather Meeker: Right … 

Irene Liu: Which is why open source AI is such a misnomer, and which is why it’s causing so much confusion, because everyone’s talking about open weights, one portion of the three different areas of what defines open source, and they’re calling that open source even though it is an open weights model.

Heather Meeker: Right. 

Irene Liu: So, can you talk about that distinction, how it came to be that open weights became open source? 

Heather Meeker: Well, I would say even people who say open weights probably are not being clear about exactly what they’re talking about, because the notion of being able to examine something, which is one of the bases of open, I don’t know that it really means that much with weights, ’cause you could get the weights, but what would you do with them? You can’t look at them and figure how they work, really. 

But another aspect of it is the rights that people get to use the weights. So, you have a model, it has a bunch of weights, and so you get that, and then the question is, do you have the right under intellectual property law or other kinds of legal regimes to actually do whatever you want with those weights?

So that is openness, not in a sense of transparency or reproducibility, but in the sense of licensing and rights. And so, I think that usually when people say open weights, what they mean is either absolute discretion to do what you want, which is more like open source licenses, or maybe just free of charge, which is how a lot of people actually use the term.

Irene Liu: And in 2023, you actually had a TED Talk on this topic about why we need open source AI. So, what has changed since then, and why is open source having its moment now? 

Heather Meeker: I think it’s a very interesting question. To me, it should’ve been having its moment all during this time, because I wasn’t the only person who was focused on the need for open source AI, whatever that means, from the beginning.

But it went off the radar for a while as the market played out in different ways. And also, the big purveyors of AI tools, and also the purveyors of the so-called open free tools, they’ve all been jockeying in the market to see who is doing better. There are elements of U.S. versus China and Europe and so forth.

And so now, I think the reason we’re talking about it more now is that people have- Figured out that the big purveyors of AI models, they’re not giving you anything, right? There’s, there’s no weights, no nothing. Most of them are just providing online services, which is kind of the most closed way to provide anything.

And then you’ve got some other purveyors who will actually let you run local copies of the AI models, and each of those things has its plus and minuses. Obviously, it’s great to have control over the computing that you’re doing. But on the other hand, I just saw an article the other day saying that open source AI is pernicious because anybody can do anything they want with it, including very dangerous things like recreating smallpox.

And so there are people on both sides of this debate, some saying it has to be provided in a controlled way, and some people saying it has to be provided in an open way. And I don’t think there’s one way that should be obviously better than the other. It’s just a very complicated question. 

Irene Liu: So currently, though, the landscape comprises of a number of companies providing, including Chinese companies, providing open weights models. Do you see any companies right now providing truly open source LLM models, or are they all just open weights? 

Heather Meeker: If we mean open source in the sense of, they are available under licenses that grant unrestricted rights, yeah, there are a number of those out there, and some of them are very interesting models and very useful models.

But I don’t really know any model that I would call fully open source because of the issue with recreating the data and providing the data. 

Irene Liu: And part of the reason a lot of the open source conversation has reached a lot of momentum is part of the Chinese models that have become so robust and have become so central to open source AI conversations.

So from a capacity and performance standpoint, how close are the Chinese models to the American models? And are these models changing the competitive landscape for the US tech companies? 

Heather Meeker: Well, you have to ask that question every day of the week because everything is changing so quickly. But yeah, there are at least a couple of Chinese models that are extremely useful, and the fact that you can run them locally is a huge benefit to people who are going to do certain things with them.

So if you ask what is a better model, some people might say anything that I have to run on someone else’s computers is not a good model for me because. And I don’t care how technically great it is because I don’t wanna give my information to someone else. So, they might prefer a model with different capacities as long as everything they’re doing with it can be kept private.

But yes, there are definitely open weights models under open source licenses that are starting to rival the big providers. The big providers in the US are, I think, almost exclusively just as a service, and they don’t provide the weights at all. 

Irene Liu: And you’ve observed, having worked in open source for a while, that oftentimes open source technology, I think you said that it, it all creeps up like a mouse in the back door, is a reference that you mentioned about how open sources creeps into organizations.

So I’m sure many of these open weights models are creeping in into organizations. And so, if an organization is thinking about governing it from a compliance standpoint, how should legal and compliance teams think about how they should allow such softwares to be used throughout the organization? Are there governance tools and mechanisms that they should be thinking about?

Heather Meeker: Well, so, open source is like the ultimate shadow IT in organizations. It has a long history of being used without corporate approval. And so, when AI came out exactly the same thing happened. It, it was eerie, as a matter of fact, because I had been working on open source compliance for years and years, and one of the big problems my clients always had was trying to get control over what their engineers were using. And companies, like, developed protocols for that, more or less. It’s not like it doesn’t creep in. It still creeps in, but they kind of got the idea about how to control that. 

When AI hit the news, exactly the same thing happened. You had employees who were using private accounts, which is a big concern for companies because the private accounts actually have different terms and conditions, and they’re probably a little less private about information.

And so, all over again, all the companies had to think, “I am asking my employees to do all this work, and I want them to use AI, but I want them to use approved AI.” So, it, it happened all over again. And this kind of thing is mostly a question of just internal IT protocols, and it is never perfect. But the important thing is to actually have a policy or set of rules instead of just ignoring it, because ignoring it is what happened originally with both open source over many years and then AI. 

Irene Liu: And with open source software, like you said, with a lot of engineers using it, oftentimes it was really hard for legal departments to get a handle of it and, but there were softwares that you could use to check before a product was released to make sure that the licenses are correct and that you don’t have any certain licenses that might release your proprietary information and whatnot.

Is there such a tool available now for AI open source, open weights, models? 

Heather Meeker: Well, first of all, the open source tools were really only as good as the engineer’s willingness to work with them, you know? 

Irene Liu: That’s true, yeah. 

Heather Meeker: And that was not always 100%. Things always crept in the back door, like you said, like the mouse tunneling in. But with AI, if you have people using AI that is an online service, that’s easier to monitor because you can tell something about what your employees, which servers they’re hitting and so forth and so on. With uh, IT policy, you know? 

[00:15:07] But if your engineer or anyone else goes home, uses their own personal computer to run some AI, either an online service or a local model, that is extremely difficult to control, and it happens all the time.

Irene Liu: So, any advice for those in-house teams other than to work with their IT teams? 

Heather Meeker: Yeah. I think that it’s important, just like it was with open source, to convey to the engineers and other people that you’re working with, “Yeah, we want you to be able to use these tools, but you have to use them in the right way.”

If you take the approach that no you, you can’t use them at all, or you have very conservative policies that makes it more likely that the people will ignore you and work around you. Like, you want legal to be viewed as a help, 

Irene Liu: Yes 

Heather Meeker: … instead of just a bottleneck, and that’s been an important challenge. And uh, what I always tell in-house lawyers who are in this situation is, “Look, you’re probably using these tools yourself, right?”

It wasn’t so much true with open source, but what I used to tell them is, “Look, you’re writing a contract. Did you write that indemnity from scratch? ‘Cause I don’t think you did.” Right? I think what you did was you took a form that you had somewhere, and you dropped it in, and you messed with it a little, and you made it work with the agreement you were writing, and that is exactly what open source is, you know? So, it was not appropriate to turn around and tell your engineers, “No, you have to write every line from scratch because of our IP concerns.” Well, you are doing exactly the same workflow. 

So it’s really more of understanding the circumstances in which your clients are working and having some empathy for what they’re doing, as opposed to putting the legal department forward as just gatekeepers and people who say no.

Irene Liu: Yeah, no, definitely. And with Chinese models in particular, though, there’s a lot of attractiveness to them because of the fact that its capabilities are increasing, and it’s becoming very robust, almost comparable to US models. And at the same time, it’s available for very cost-effective- 

Heather Meeker: Yes … 

Irene Liu: ways. And so, for companies that are evaluating using these Chinese models within their companies, are there additional security data governance considerations that one should be thinking about?

Obviously, even if you are a department of yes, I’m sure you still need to think through some of these considerations. So, are there additional considerations for Chinese models versus US models that are open weights that you should be thinking about? 

Heather Meeker: Well, so, when people say they’re using Chinese models, I think mostly what they’re talking about is running them locally.

Irene Liu: Yes. 

Heather Meeker: And that is actually a little bit less risky than using a service. I think if we were talking about Chinese models that were run out of China, the almost the universal answer would be, “No, we don’t want you doing that because of security and secrecy concerns.” So, they’re really two different animals.

And certainly, if the model is properly sequestered on your own computers and there are no backdoors that it’s going to, then great. It’s just that as the models and the agents that they run… I mean, AI these days is not just a model. It’s like this very complex network of functionality that uses a lot of different elements, and you need to make sure that the models you’re using are not reaching out to other things that you don’t have control of.

And honestly, I don’t know whether it’s really possible to control all of that, but that is definitely what IT security people are focusing on now. 

Irene Liu: So legal departments just need to partner very closely with IT teams in general. 

Heather Meeker: They do. Yeah. And this stuff is genuinely hard to keep up with. It literally is changing every week, every month and it is getting more complex every week, every month.

And the compliance tools that are available, there are many, many startups right now, I know from my law practice, many startups who are working on things like providing guardrails and filtering prompts and responses and keeping track of agents that AI is using. And so those tools are all catching up, but they’re always playing catch-up, you know, so it’s never gonna be perfect.

But I would say you should be focused on and keeping up with what the, the tools are as well, and they’re gonna develop very rapidly, and it’s gonna take a lot of effort to keep up with them. 

Irene Liu: Yeah, and it sounds like a partnership that needs to deepen between the legal and IT teams because frankly, it’s always been a collaboration between the two teams, and there’s always been security issues, data privacy issues at companies that the GC and the office of the general counsel often had to think about with the IT teams.

But now with AI, it seems like they have to deepen their collaboration even further and also stay abreast of all of these AI developments as well as the tools that might be available to help them with these guardrails. 

Heather Meeker: Yeah, definitely. I went to a meeting a while ago that was arranged by a private company, just hosted it, of a lot of lawyers who were using AI to discuss how they were using AI, and that was extremely helpful, you know, because I think a lot of the lawyers feel isolated, and they don’t really know what’s going on, and they have trouble keeping up with all of it. They have to keep up with all the legal stuff too, right? 

So I would say talk to everyone you know about what they’re doing because we all have to be sharing best practices at this point even more than we used to. 

Irene Liu: And talk to everyone within the company that you know. Is that- 

Heather Meeker: Within the company- 

Irene Liu: Or, and outside.

Heather Meeker: And also, outside. Most lawyers have lawyer friends, and they’re all struggling to figure out How do I use this? Should I use it? How do I use it safely? And you can learn a lot just by having a short conversation with people. 

Irene Liu: Yeah, most definitely. And the other interesting thing about open source is, and open weights in general, is that there’s been a lot of community building around that recently.

There was a letter, the Open Weights in American AI Leadership open letter, that was signed by over 230 companies that urged policymakers to avoid premature bans and restrictions on open weight models. What’s your reaction to that letter? 

Heather Meeker: Well, I might not be objective about this, but I think open is probably a good way to go. It’s just a question of how to execute on it, and I’m really glad… 

Like, it would be a mistake for governments to say, “Open models are bad because we don’t control them.” Well, I guess if you are a person who believes government should control everything, maybe you would disagree. But my own personal opinion is that having the government control this in any, you know, really fine-tuned and significant way is really not gonna be the right outcome. It’s going to hamper technology. It’s also got inherent risks of government surveillance and government control where we don’t want government control. 

So, I’m really glad that the people who are in favor of open, their voice is being heard more because this is the first time I’ve really been hearing a lot of people get behind this idea, even though we were talking about it two, three years ago. So that makes me happy. But there are definitely people in government, people in, just in general walks of life, who are really concerned about having open technology available because it’s not controlled. 

So the question is, do you believe that it should be controlled centrally or not? And I don’t think there’s one answer to that question. I would probably lean on the side of no. 

Irene Liu: And just taking that a bit further, so with an open source, if it’s truly open source, there’s transparency elements to it, but if it’s open weights, there are potential safety risks because of the fact that you really don’t know the source code and you don’t know the inputs. There’s a lot that you just don’t know with open weights models. 

Heather Meeker: Yeah. I, I would say that’s true with either open or not open. It’s an inherent issue with AI. And we all know from reading the news that even the brilliant people who are working on the most cutting-edge models are not always sure what their models are doing.

Irene Liu: Yeah. And we saw that recently- 

Heather Meeker: We did 

Irene Liu: … that there were agents that were doing things that were totally unauthorized and on their- 

Heather Meeker: Yeah 

Irene Liu: … and rogue behavior. So, is your opinion that open source AI is the- Not any safer or riskier than proprietary AI, and that it gets a bad reputation to a certain degree? 

Heather Meeker: I would agree with that. Maybe it’s because generally I’ve seen open source software get blamed for a whole bunch of things that it really never should’ve been blamed for. 

There’s a tendency of people to say, “Oh, this problem happened with an open thing, so open is bad.” That was always part of a security debate, although I would kind of put debate in quotation marks because I think most people in the technology business understood that open source was actually more secure than proprietary software.

And you’re seeing that kind of dialogue come up again, where people are saying, “Well, open source is risky.” Well, I don’t think it’s the open nature of it that makes it risky. I think it’s the uncontrolled nature of it that’s risky. And we have to make a societal decision about that. But transparency is almost always something that is good to control behavior. So, I think open is the solution. We just have to figure out what it looks like. 

Irene Liu: I think you mentioned just earlier that really at the end of the day, the open weights models are much safer or riskier than proprietary AI because you really don’t know what’s happening behind either of them. 

Heather Meeker: I don’t think it’s inherently riskier.

Irene Liu: And is it also because open source typically has gotten a bad reputation, like you were saying, because of the virtue that it was open, but you also like the fact that it is open and there’s a community that can help spot these safety risks? 

Heather Meeker: Yes, that’s exactly right. In open source, there’s something that’s informally called Linus’s law, which is, with enough eyeballs, all bugs are shallow. It– I think it’s completely apocryphal, by the way, and he, he may have never really said it. But it is a really interesting idea. It’s like sunshine is the best disinfectant, and that’s a really important idea in open. 

I think with AI, the challenge is, how do we achieve that kind of transparency? If we can achieve it, then that will be the safest kind of AI that there is.

Part of it is disclosing how models were trained, because if you can look at the training methodology and say, “Well, this was a biased kind of training in the scientific sense, in the social sense,” whatever, then you can say, “Okay, that transparency allowed us to figure out whether this was a good model or not.”

At this point, the technology of AI really is kind of black box technology, but I would not sit here and think that will always be true. In fact, I know that there are people who are working on ways to try to unpack AI models, and that could happen tomorrow, and it could change everything, you know? So, the more technical possibility of transparency, I think the stronger an open model looks, because then you really can look at it and say, “Okay, I can look at this thing, and I can know that it’s not doing something nefarious that I don’t want.”

And that’s the way it worked with open source software, and I’m hoping that’s the way it will someday work with open AI. 

Irene Liu: And when we’re thinking about open source AI, what is the current landscape from a regulatory standpoint? How are the US and the EU grappling with open source AI and open weights AI?

Heather Meeker: Well, there is federal legislation that’s been discussed. We don’t have a cohesive regulatory system. Of course, that means that we’re likely to get a crazy quilt of state regulations which may not be all that useful either. So, we don’t really have anything cohesive in the US.

In the EU, there is a new AI regulation which has just come into effect, but it is based on the nature of use of the AI rather than the AI itself, and it requires- varying levels of disclosure depending on the use of the AI. And that kind of overlaps with the idea of open or open source, but it’s definitely not the same thing. 

So, the EU has tried to do that, and we’ll see how it works. With all regulations, they tend to be, well, for one thing, often written without maybe enough attention to technical realities, but also, they can be very difficult to adhere to in practice. And I’m sure that we will spend the next few years figuring out how to adhere to the EU regulations, because it’s just come online. And in the US, we’ll just see what happens, see if our government can focus on this enough to do something sensible 

Irene Liu: Well, most recently, the White House announced a new framework for regulating AI models, but open weights models were explicitly excluded from it, as far as we know. These are secret discussions behind doors. So, we know that there’s a new framework, and we know that open weights models are excluded. We don’t know further details beyond that. 

But for those open weight exclusions, what are your thoughts on that? Do you expect regulators to continue to treat open and proprietary models differently?

Heather Meeker: I would hope that they would. The main thing I hope for, and I don’t know if the framework will result in any of this, is that anything that is adopted by government should be open. This is a very old idea in open source, that if a government chooses to use something as opposed to a private actor, there should be enhanced levels of transparency involved. Because the government is doing things that is a taxpayer expense, that affects all our lives, and should have notions of due process and fairness, and those ideas are probably best served by open models. 

So what I would love to see is that the government says, “If we’re gonna use a model, it has to be open according to whatever definition they come up with,” as opposed to the government trying to regulate all models. We’ll see what happens there. I don’t think it’s very easy to predict at this point, but I would love to see requirements of openness for government use of models. 

Irene Liu: And Heather, you just released a new book. Can you tell us about that new book? It sounded really interesting. 

Heather Meeker: Yeah. A friend of mine and I wrote a book called Open Source for Babies. This is, of course, for a somewhat younger audience than I’m accustomed to writing for. 

Irene Liu: Who is it directed to? 

Heather Meeker: Yes. My friend, he and his wife just had a baby, and he said to me, “We need to write Open Source for Babies,” and I’m like, “That’s a great idea.” 

And so we came up with a book which, maybe not so much due to us, but due to the illustrator who did an amazing job, is a very cute and sweet book about teaching small children the value of sharing, right?

We’re having a book release party that’s coming up pretty soon, and my friend has advised me that he read the book to his infant son who enjoyed it very much. 

Irene Liu: Oh, that’s amazing. So, what are the other elements for babies that they should know about for open source other than sharing?

Heather Meeker: Well, in the book we used, as a metaphor, the idea of making music. So, the idea was you can make music on your own, but if you make music with your friends, it’s more beautiful. 

Irene Liu: Oh, that’s awesome. 

Heather Meeker: So that, that’s the idea. That, that is definitely an age appropriate idea. There are other books out there about open source for young children, but we were really going for – 

Irene Liu: Yeah, babies

Heather Meeker: the babies. Right. 

Irene Liu: But that’s such a great spirit of open source, and so- 

Heather Meeker: Yeah 

Irene Liu: … that’s definitely something that I’m sure a lot of babies will get, hopefully, from the book. 

Heather Meeker: Uh, it’s something I believe in a lot in all aspects of life, and I have to say that I’ve tried to also conduct my professional career in a way that involves maximum sharing. because I think that’s actually, number one, it makes your life better, number two, it makes other people’s lives better, and number three, it actually makes you successful. 

Suboptimizing is not the way to succeed. The way to succeed is to share your knowledge and help other people, and when you do that, that’s actually a real contributor to success and also to just happiness in what you do for a living. And so, I, I actually believe in that deeply. 

Irene Liu: And I love that spirit ’cause I’ve also benefited from your book, Open Source for Business, when I was also in-house, and so that book was really instrumental for me to understand the open source ecosystem, so thank you for sharing that. And now thank you for writing such a fun book for babies. I’m sure babies will appreciate it.

But looking ahead, like, what do you expect the open source AI ecosystem to look? Are there things that we should be doing to prepare for that future as lawyers? 

Heather Meeker: I think for lawyers at this point, the main thing to do is just to be keeping up on what’s going on. And I’ve been at this for decades, this technology law stuff, and I’ve never experienced anything quite like this in terms of how fast things are changing. And the real difficulty in keeping up with stuff is filtering out all the noise. 

Irene Liu: Yes. 

Heather Meeker: ‘Cause every, every day you open up your newsfeed, and there are, like, a ton of articles that are just garbage and not interesting. So, I would invite people to find a few commentators who they think are both smart and able to explain things well, and follow those people, and maybe check in every month or so.

Maybe not every day, because every day is overwhelming, right? But check in every once in a while, to make sure that you’re keeping up on the latest developments so that you can just understand, at least at a high level, what is going on with this field, because it is changing so much every single day. 

Irene Liu: And how are you keeping up, Heather? Are there things that you would recommend in particular? 

Heather Meeker: I look at a number of newsfeeds. I get very frustrated with the news because I think a lot of it is really low quality. I’m sure I’m not alone in that. But yeah, you gotta look at the sources that you trust and that are actually useful to you. 

A lot of news about AI is written to scare you or tell you how great it is, and so I’m not interested in either of those things because those aren’t useful to me. What I wanna know is, like, what is actually happening in technology and in business, and I want the facts. I don’t particularly want what people are gonna, like, tell me their opinions about it. That’s just me. 

Irene Liu: As a final takeaway, are there one or two sources that you would refer our listeners to? Obviously to the AI Sidebar, but anything else that people should be utilizing that you find to be a great resource for you? 

Heather Meeker: I will betray my predilections, I guess. Sources that I like, Wall Street Journal technology section- Yeah … I think does very good reporting. Economist technology also does very good reporting. Wired is usually a good source. TechCrunch is usually a good source. Those are more like the Silicon Valley sources. 

Irene Liu: Sure. 

Heather Meeker: I don’t usually spend a lot of time on the kind of more general news sources that are not business related because I find what they say too shallow and not useful. 

But, I’m very interested in the tech business, so that’s why I tend towards the business related publications. I wanna know who’s making what, are they making money at it? Is there any AI business model that’s sustainable? By the way, we didn’t talk about that, but that’s a really big question. And so, I’m focusing on a lot of that. 

And then I’m also looking at, like for instance, the LinkedIn news source is pretty good, but you have to filter, like, to find out what people are doing with AI. It can be useful, particularly if they’re people I know. So, I’m always asking people, “What are you doing with it?” You know, particularly the lawyers, to try to find out, like, how is this actually useful to you, and how do you use it on a day-to-day basis? 

Irene Liu: Yeah. Well, it sounds like canvassing a lot of news articles from reputable- 

Heather Meeker: Yeah

Irene Liu: sources and talking to people in general, so. 

Heather Meeker: Yeah, definitely. But it’s a job absorbing all the information that gets poured out at us in this society, and you gotta focus on what is actually useful to you. 

Irene Liu: Yeah. Well, thank you so much for that takeaway, Heather. I really appreciate you being here and for sharing your insights.

I know you are a reputable expert in this topic, so I’m so grateful that you’re here. Thank you.

Heather Meeker: Oh, thank you. This was lots of fun.

Irene Liu: A huge thank you to Heather Meeker for joining us and helping us unpack what open source AI really means, why it matters, and what legal and business leaders should be thinking about as this landscape continues to evolve. Like we discussed, Heather has recently published a children’s book, Open Source for Babies.

The book uses music to introduce the ideas behind open source, including sharing, collaboration, and building something better together. It is a creative reminder that some of the most important ideas in technology can also be explained in simple and accessible ways, even for babies. And thank you for tuning into The AI Sidebar.

If you found value in today’s conversation, follow the podcast and share this episode with someone who will benefit from it. We’ll be back in two weeks with another thoughtful conversation at the intersection of AI and law. So, until next time, stay curious and keep learning.