Protege on the Future of Licensing, Content Rights and AI-Driven Media

[MUSIC] You're listening to Inside Content, the TV industry podcast. This show is brought to you by Three Vision, a global TV industry consultancy specializing in content acquisition, strategy, research and business development. Each episode we give you VIP access to the views and experiences of senior TV executives and discuss the latest TV industry trends and insights. Hello and welcome to Inside Content. I'm Eva Demora, analyst at Three Vision, and today I'm joined by Dave Davis, Chief Content Officer at Prodigy. With a background spanning roles at Paramount, NBC Universal and False, Dave now leads Prodigy's efforts to license media content for AI model training. In this episode, we unpack how Prodigy is bridging the gap between content owners and AI developers, the challenges of licensing in a rapidly changing landscape, and what makes her content more valuable for training. Dave also shares insights on the growing demand for curated data sets, the impact of metadata, and how thoughtful licensing can help content owners protect their rights and shape the future of AI. Let's dive in. Introducing Video Markets Tracker from Three Vision. Your go-to source for global TV data. We offer regularly updated regional and global reports covering Pay TV, S-Vod, A-Vod and Fast across Africa, Mina, the Americas, Europe and Asia Pacific. Makes smarter decisions with insights you can trust. Learn more at threevision.tv. Now on to the episode. Hi Dave and welcome to Inside Content. How are you doing today? I'm good, thank you. Thank you so much for having me. Thank you for coming on. I'm really, really excited to speak to you today about the world of content licensing for AI models and for training. But before we jump into that, I want to give our viewers and our listeners a bit of a chance to get to know you a bit better. So I want to ask you, what's your favorite film of all time? So I'm a long time, I was a long time paramount executive, so I should say the Godfather, and I love the Godfather, but I'm going to say the Shawshank Redemption. Amazing, amazing film, very good choice, amazing. Well, now let's not waste any more time, shall we? I want to ask you, can you tell us about your current role at Prodigy, and your journey from traditional media licensing to the world of AI training data? Sure, so I'm Chief Content Officer at Prodigy. We, so Prodigy, just for a quick background, we're a content aggregator. We aggregate content from content owners to licensed AI companies to use for model training. How did I get here? It's been a long journey. I started a million years ago on Wall Street, actually, at Goldman Sachs, and then I went and practiced law actually for a couple of years. I didn't love being a lawyer, and fast forward a couple of years from there, moved to Los Angeles, spent time at, started an entertainment industry career at the Motion Picture Association, and then was an executive at NBC Universal and their International Television Distribution Group, and then on to Paramount, where I did a number of things at Paramount over nine years, but primarily managed their international VOD business, and then was recruited to join Fox after that, and also ran their international VOD business right up until, and a little bit after the Disney acquisition, and then had an interesting role as Chief Commercial Officer at Motion Picture Licensing Corporation, which was a private equity back to company. In that role, AI started to, you know, chat GPT launched, and all these interesting things were happening, and I became very enamored of it. I became, you know, I'll call it an early adopter, but in the way, you know, sort of doing parlor tricks at the bar showing people all the cool things that chat GPT can do. And started to become aware of some of the copyright issues around that, and thought that the world was going to evolve to a place where copyrighted works would be licensed for model training. So, launched a company called Colliapy Networks, long story short, in December of 2024, Colliapy Networks was acquired by protégé, which is a really phenomenal startup, amazingly talented team, and, you know, we're often running. So, protégé actually started in the healthcare data space, but their vision was always to provide data across verticals to AI companies for training. So, they were already looking at the media space when they met me, met my co-founder, and we've come aboard. It's been an amazing journey. We now, I mentioned we now worked on behalf of about 100 media companies across six continents. I'm very proud to say that we've paid down millions of dollars in royalties this year. And, you know, we're trying to do this in a really thoughtful way that keeps all of content owners demands very much front of mind. So, you know, I've spent most of my career, again, on the distribution side. So, I have a pretty strong understanding of the concerns that media companies may have when they enter this exciting but scary new world. That's amazing. That's really, really cool. I mean, you've had such an impressive career, and it's amazing that you were such a visionary. You saw the power of AI and the potential, and just jumped into it. It's really, really cool. Can you tell us a bit more about protégé? I mean, you've mentioned they started in a different sector than they had moved into media when they acquired Kalaivevi networks. Can you tell us a bit more about the companies? Yeah, the company is, let's see, what to say. We're on a great journey. We're on a great growth trend, both the healthcare vertical and the media vertical and then our newer verticals in audio and motion capture are really often running. Yeah, I think there's a lot of our track record that I can't tell you because it's confidential, but I can tell you that based on our track record, we recently closed a series of financing for $25 million in August. That sets us up well to continue to invest in technology and in superstar talent really to do the work, the good work on behalf of content owners around the world. That's amazing. Well, congratulations. That's super exciting. Thank you. And looking at the future, is there any other sector that your company, that you can share, that your company's looking at, or are you going to focus on what you're currently doing in those areas in particular? So I'm going to be focused on those sort of media-related verticals. So audio, visual, audio and motion capture. And then we will enter more verticals, but yeah, nothing that I can disclose. That's completely fair. Well, thank you for that. How does Prodigy Media's approach differ from traditional content licensing models? So in some ways, we're, we don't differ that much, and in some ways, we're a traditional aggregator. We work with our partners on a revenue share basis, our partners take the, we license content from them and then sub-license it to AI companies, and then share the revenue on an agreed revenue share basis. So from that standpoint, we're not terribly different from plenty of other aggregators out there. In other ways, we're very different. This type of license is very new. I mentioned it's, you know, there's a lot of sort of chaos in the market that we're working through. There's the then diagrams of what content owners want and what AI companies want. Sometimes don't overlap initially. We talk about that one of our big roles is really to be a translator between those worlds. I mean, at a simple level, you know, in the content space in the media world, we all call, you know, a movie or a TV show. A movie or a TV show. AI companies call it data, and then you take that and magnify that into legal contract language and so on. There's all kinds of words and terms that don't just don't mean the same thing in those worlds and provisions that end up being pretty difficult negotiations. So these deals aren't easy, but we think, you know, we think we're one of our, the data, just really, is that we've been able to go to market. We've been able to persuade AI companies of a lot of the commercial practices in media that aren't worth their time to fight to change. Such as, you know, a few months, six months ago, a year ago, most AI companies wanted perpetual rights. On the media side, you know, most people don't have perpetual rights. So, you know, so that that was a really difficult issue, you know, six to 12 months ago. It's becoming much less difficult as we've had a significant role in persuading AI companies that that's not a hill worth dying on to them. The other thing I'll say is unlike sort of a traditional film aggregate, we do a lot of, there's a lot of technology that goes into this. So we spent a lot of money on metadata related to content at the frame level. So not even like title level metadata. We have a product called SHOT that we've invested quite a bit in that allows us to, if an AI company, some AI companies want to just a license of a full length movie or a full length series or news or whatever sports, but others really want much more curated content. So they might want just a thousand shots of a horse walking across a meadow or a thousand car crashes or whatever it is. And we can, we can do that, you know, now pretty easily because we've developed a lot of technology both in-house and working with partners. That's amazing. I mean, I can imagine it's, it's really, really cool to be in the middle of it all. And with your background, I mean, background in law, background in media, you must be the perfect person to kind of be the messenger and the translator, like you said. I say, I said, you're trying to say that. Thank you. Of course. But I can imagine, yeah, it can be a lot to translate for these companies because like you said, AI companies want something quite specific at times. And I can imagine content owners, you know, the prospect of AI and AI and licensing their content for AI is one that's been talked about a lot. There's a lot going on within. So how do you balance the interests of content owners with the needs of AI developers? How do you make sure that everyone's getting what they need and everyone's kind of protecting their interests? Yeah, I mean, so the first comment is that, you know, I come from the content world and a lot of my team comes from the content world. So, you know, so I worked at three studios. You know, we've got people who worked at, you know, studios and media companies that I didn't work for. So it's pretty broad content experience. So we're very cognizant of the concerns of content owners. And that can be things like, you know, content owners don't want us to give an AI company distribution rights, right? They don't want them to have the right to publicly display content, you know, in their outputs, exact copies of what's gone into the models. They're going to have concerns about, do they have rights to license music? You know, probably not. So we probably need to carve that out. Do they have rights to license the face or the voice of talent, you know, probably not unless that talent is given an express waiver that is an express release that they can do that. So we need to be very careful and explain to the, you know, AI companies sort of what the content side of the world can do and and what's less possible. But having said that, we're a market maker and we have deep relationships with the AI side as well. And we've developed a very good understanding of what AI companies need, what they want. And we do spend a lot of time explaining to content owners, listen, this deal is going to be a little different than your other deals. And you do have, if you want to be in this business, you're going to have to concede a point around, you know, at Swiresy. You know, one, one example of that is, you know, while we've fought back perpetual license periods, you know, license periods can be, can be long in some deals and hopefully the structure and the compensation, you know, makes up for that. So, you know, to the point of what AI companies need, also there's just, there's specific kinds of content that they'll want, there's specific kinds of shock curation that they may need and we've become, you know, pretty pretty, that's pretty good experts at that as well. Yeah, yeah, that makes a lot of sense. I can imagine each AI company has a different goal, different model and they're looking for different stuff. But in general, are there any types of content that are particularly useful or valuable for AI companies to have and indeed types that aren't really, they don't really seek out very often? Yeah, so it's interesting. So, you know, there's, there's, first of all, there's all kinds of different AI companies, right? So, there's, there's generative video companies building foundational models and they may want, you know, kind of everything. And there's also companies that are focused on, you know, avatars or focused on the effects or on editing suites. And then there's even, you know, AI companies that we talk to who aren't in video at all. They're in, you know, they're doing things around robotics and using video to train, you know, a robot, how to, you know, pick up a pen or whatever it is. So, there is a, there's a broad diversity on what different models need. But in general, if we're talking about generative video, they want, they tend to want crisp video with motion, well lit, steady camera. Camera movement is good, but not shaky camera. They will often say they want cinematic. And for the most part, they do want cinematic, but they don't want all cinematic. So, for example, we've learned a bit horror content can often not be great for a model because it is, you know, there's a lot of scenes that are really low light to build tension and there's a lot of scenes with a, you know, handheld camera that was purposely really shaky to build tension. So, that doesn't work very well for model training. You know, the things that don't work, they don't love subtitles, they don't love graphics. We occasionally get the question, could we, you know, we don't want, you know, somebody's face to appear in a model, can we blur the face? Really, can't blur a face because it's going to, it's going to teach a model. The model is very literal and it's going to teach a model that a human face is blurred. So, you know, so that, that becomes difficult. Animation is very much in demand. And then kind of everything else, we sort of joke that every time we think there's something that an AI company, that no AI company will want, you know, three weeks later we get a request for it. So, to broadly speaking, you know, we get requests for news, for sports, for, you know, for fashion, for cooking, for travel, wildlife, you know, on and on. It's sort of, you know, there's a buyer for almost anything. That makes a lot of sense. I mean, hearing you say that horror films are not very good. It makes sense. I mean, the whole premise of it is that the video is not clear. That's how they build it. I'm surprised to say that there's a lot of demand for animation, though. I'm surprised to hear that. I wouldn't have expected it. But yeah, that's really interesting. Sorry, go on. Yeah, I think there's part of it is just scarcity. There is less animation produced and, and, and less animation that's been made available to, to AI developers. That makes sense. And, um, do you think it doesn't make a difference at all to the content owner? What sort of model their content is being used for? So in the conversations that we've had with partners, not so, not so much, I could imagine a content owner saying, you know, what, based on some concerns that people share, you know, a lot of content owners are, are worried that they're sort of feeding their demise or feeding a competitor. And I suppose that for those, maybe they would be less concerned about, you know, or robotics model, because it's not really relevant to media. But the generally, that isn't a conversation that they don't have to anybody that I can remember. I mean, you've mentioned that animation, for example, was very valuable because there's not a lot of it. Apart from the exclusivity or the diversity that there is, is there anything else that makes the data said valuable to AI developers like metadata richness or anything like that? So metadata richness is definitely valuable. And sometimes, you know, sort of ancillary content related to the video data, you know, so we will occasionally get requests for transcripts or for, even, you know, for screenplays. Diversity of content overall is valuable. People don't want a model that is, you know, biased, you know, ethnically or locations or anything else. So the fact that we have partners on six continents is a selling point. And then just probably speaking, you know, content type, what we talked a little bit of as genres, quality and resolution. So most models want to train on the pure, you know, pure HD or higher resolution. You know, and I think that's going to continue to evolve. We're starting to get requests for 4k. That's, you know, we expect that's going to be more and more at some point. I think AI companies are going to start asking for, you know, HDR or the old division or things like that. That makes sense. And can you share some insights on the current state of the AI training data market, particularly in the media? Yeah. Um, I'll say it's chaotic, but, uh, but actually substantially less chaotic than it was six months, six months ago, three months ago. There is, there's, you know, there are some, you know, emerging, what I'll call with, with broad parameters, some market terms. The overall growth of the businesses is really strong of the market. There's, there's a lot of licensing deals happening, mostly not announced, but there's a lot happening. I mean, yeah, and it's not just protege. There's media companies doing deals directly. There's, you know, there are other aggregators. Um, you know, and I mentioned earlier, we're very proud of the fact that we've, we've paid out, you know, a lot of money to our partners so far this year and we see that growing. Yeah, that makes a lot of sense. I mean, yeah, the world of AI is so rapidly changing in terms of sometimes the legal aspect of it can keep up quite, um, at the same pace. So I can imagine, yeah, it's a lot going on. A lot of factors, a lot of players involved, and yeah, I can imagine chaotic is probably the best way to do this. Um, and how do you see the demand evolving for, for high quality license data, shaping the future development of generative AI models? So we're very bullish on it. Um, that these, these models have a pretty ravenous appetite for content. I think, you know, it's, it's the way they take the content is likely to evolve. Um, but we see them needing an awful lot, you know, when, when they need a, you know, I mentioned, this is an example that nobody's asked, I actually asked for, but, you know, if it was, you know, horses walking through a meadow, you know, they wouldn't need six shots of that. They'd need, you know, maybe 20,000 shots of that, you know, so, so, so they want a lot. If they want, you know, hours of video, it's not 100 hours, it's 100,000 hours, you know, it can be, can be more. We've talked about deals that are significantly larger volume than that. But the evolution will probably go from just, you know, massive numbers of hours to more curation as these models get better at certain things, but realize they have certain weaknesses. There will be more, uh, more curation that's, that's valuable. And in curation comes, it, it takes technology to get there and it, and it's rare and it comes with a price premium. So that's great. And, um, I mean, we've talked about this a little bit, but I do want to ask, what objections are constraints do you usually get from content owners when, when it comes to AI licensing? So, I think that the biggest objection is just, you know, broadly, well described as, as, you know, competitive fear or, uh, that, that they're kind of feeding this beast that is going to disrupt their business. Um, and, and my answer to them is usually that, that AI is inevitable and whether they, and it's, it's, yeah, it's going to be disruptive for a lot of businesses in certain ways. And, and we can predict some of those and we can't predict some of those. But none of us is big enough on our own to sort of alter that trajectory. Like it's common. Like, an AI is a little bit like, I think of, you know, thinking cinematically, I think about the boulder in the Indiana Jones, you know, movie, the boulder that's rolling out of the cave, like the boulder's coming. And, and, you know, indie can't turn it to the left or turn it to the right. So you just, you have to prepare for that boulder. So, so that sort of leaves a content owner with three choices. They can ignore it and, and ignoring the boulder seems like a bad idea. They can, you know, in some cases, they could litigate, uh, you know, that's, that's an option for some. What we do is, you know, we, we position ourselves in the third bucket, which is the third choice, which is engagement. So we think there's an opportunity to sort of shape the world through licensing and, uh, you know, and make it about a world for content owners, but, but, you know, in living in the real world and, and not ignoring that, that this is a disruptive threat that's coming. Yeah. Yeah. I think that makes a lot of sense. I mean, yeah. Like you've mentioned, it, it's coming and it's happening. And with the, I mean, you've mentioned litigation. But the recent lawsuits that have been going on and discussions and controversies, have you seen a change in either the attitude or the volume of demand for content on the side of AI companies or of content owners that are willing to license their content? Yeah. So I can't necessarily say that it's directly attributable to litigation and, and, and, you know, like the day, Gantz-Ropp, the copyright settlement in the US, the big, you know, the largest, uh, copyright settlement in, in US, uh, history, uh, 1.5 billion dollars. But, but there's no question that, uh, um, there is, there is a lot more activity. And I think it's a reasonable hypothesis that, uh, that Sunday, AI companies have decided that, you know, licensing is a, is a, is a good path to go down. You know, we have, we have really strong engagement from most of the largest models and, and many medium size and small models. And we see a pipeline, by the way, of like, this isn't, uh, a world where there's only five or six customers. This is a world where there's probably 50 customers per unit to 500. That's really, really, really interesting to hear. And we've talked about the objections or concerns. What potential benefits do you see for content owners who choose to license their content? Yeah. I think there's a couple of big things. One is monetization. This gives them an opportunity to, you know, have a purely incremental revenue stream. So in a world where a lot of my friends who are still in content licensing roles are seeing their revenue go down. This gives them a path to revenue that is, that is totally incremental. You're licensing to data scientists sitting in a computer lab somewhere. Um, you know, so, so it's, it doesn't put any mileage on the product. But the other, you know, maybe more abstract benefit is, um, that you get to shape this world. The, uh, this, the CEO of the Atlantic, when they did their, they did a news deal with open AI and, and the Atlantic came under some criticism for that. And the CEO of the Atlantic, you know, said, basically, you know, I don't know how this is going to work out. But I feel like we have a better chance to shape this world. If we participate in it, then if we ignore it. And I think that's exactly right. Along those lines by merely participating and making content available to AI companies for model training, we strengthen copyright. So under at least under US law, some of many of your listeners have heard a lot of talking about the fair use provisions under US law. I mean, fair use is a four factor test. And part of the test for one of those factors is whether content has can be commercially obtained. So if AI companies, if content owners don't make content available at all, it makes it much more likely that, that AI companies at least under US law can win a claim that they just, it would be impossible to license this. So it should be fair use. That makes sense. So it's actually protecting their content in a way as well. That's right. That's really, it's really, really interesting to hear. And how, how is footage and media working to expand its network and find new content providers? Part of it is through participating in panels and educational outreach. So appearing on this podcast is one way. You've got a great audience that will hopefully be interested in this topic. I spoke at MIPCOM last year. I spoke at Content Lundin, the television academy, and then going forward, I'm speaking at MIPCOM again in a few days at Asia Television Forum, potentially. And likely at CES for digital Hollywood. The other thing is, you know, we've got a partnerships team. I'm part of that partnerships team. And we spend a lot of time taking inbounding queries and reaching out to companies that are interested in AI licensing. That makes sense. You've mentioned that you work with bigger companies as well as medium-sized, small companies. Do you see it different in how the dealing is with, well, depending on the size of the company? I think what they're looking to buy is often shaped by the size. So the largest companies may be more interested in time age and a ton of content. They may be really interested in some specific curated things as well. But the smaller companies may not be able to afford it, or may want to be a little more cautious and sort of carefully choose what they're going to license. But the other thing I'll say is that many of the smaller companies, some of these smaller companies have raised hundreds of millions of dollars in venture funding, and so they're not so small. But they may be doing things that are more specific. So I mentioned earlier that there are companies doing things like specific VFX or editing tasks. You know, another big one is lip syncing of dubs. So those become very specific opportunities. You know, you can imagine a company doing lip syncing for dubs. They want lots of conversational faces with mouth-moving. They would love the video of this podcast. That makes that makes sense. Yeah. And what about when it comes to content owners? Do you see a different one? It's a small creator versus a bigger studio? That's an interesting question. I think the question for all of them tends to come down to, you know, do we want to make everything available? Are there certain things that to the extent we're a little bit nervous? Maybe we're going to hold back our prized jewels, but we'll license a little bit of the deeper catalog. So I guess broadly speaking, no, but it's the nature of the content they have available. Okay, that makes sense. And you've mentioned you're going to be a MIPCOM. What are your thoughts on the event? And it's significance right now with the AI-driven media innovation? Yeah, I think it's really important. I think it's, you know, obviously from my perspective, you know, it's interesting to have a chance to talk to my former friends and colleague, or my current friends and former colleagues, about, you know, an incremental revenue stream, and in a time when a lot of them are facing revenue shortfalls. So it, you know, so as the distribution community comes together to think about, you know, how to rebound from, from the last few years, which have been pretty challenging. I think we have, we have one answer for them. It's not the only answer for them. It's also going to be a lot of discussion, you know, at a lot of these events, including MIPCOM, about the potential for, for AI to be used effectively on the production side. That's not my world, but it's really interesting world. And it's a controversial world. One thing I'll say, you know, just to make the point is that I was in an event, I spoke in an event this spring called AI on the lot in Los Angeles. And some of the coverage coming out of that event, I thought was really good. One of the trades commented that, you know, the difference between last year and this year is last year people were talking about making AI movies. This year people were talking about making movies with AI. Sort of an acknowledgement that this is, these are tools that can go into the production workflow somewhat in the same way that, you know, CGI was introduced in the production workflow a long time ago. So that doesn't change that it's controversial. It doesn't change as we said earlier that it's going to be disruptive sometimes in really exciting ways and sometimes in ways that are that are really kind of sad and have a human cost. Yeah, yeah, you're absolutely right. I was at IBC last month and I know they had a whole whole about production and it was mostly AI driven companies and how AI is completely revolutionized in the way it's done. And like you said, it is definitely going to be disruptive. It is an exciting time. I'll be it a bit scary, you may be at times, but this, this is all really, really exciting. But yeah, this has been, this has been super interesting, super informative. So thank you again, Dave, for coming on our podcast today. I don't know if there's any last remarks you want to make before we finish. Nick, you know, for anybody who's interested in contacting us, I'm on LinkedIn. You know, please send me a message and I'll be I'll give a shout out to MIPCOM. I'll be speaking in the innovation lab on Monday morning. If anybody wants to stop by everybody's MIPCOM schedules get very, very busy. But if you have time, please come see us. See me. But this has been a pleasure. So thank you so much for taking on this topic and for inviting me to speak. Thank you, Dave. This has been amazing. And we'll speak soon. Okay. Thank you. Thank you. Thanks for listening to this episode of Inside Content. The TV industry podcast brought to you by 3Vision. With decades of TV industry experience and real world success, we know the ins and outs of the market to let nobody else. To learn more about our TV consultancy services, head to 3vision.tv.