Julián Posada on the People Powering AI and the Platforms Exploiting Them
Justin Hendrix / Nov 22, 2026Audio of this conversation is available via your favorite podcast service.
Behind AI systems that can recognize a face or flag a hateful post is a person who labeled the data that trained it. Many of those people are doing invisible labor on platforms that pay by the task.
In Platform Extractivism: Data Work and the People Powering Artificial Intelligence, published this fall by University of California Press, Yale University assistant professor of American studies Julián Posada follows the Venezuelans who became some of the industry's most important data workers during the country's economic collapse and the COVID-19 pandemic. He argues that AI's promise of freedom, both to work and from work, has yet to arrive for most people, as gig-style terms spread even to doctors and lawyers. I had a chance to speak to Julian about the book.
Julián Posada:
Hello, I'm Julian Posada. I'm Assistant Professor of American Studies at Yale University, and I'm the author of Platform Extractivism: Data Work and the People Powering Artificial Intelligence published this year in 2026 by University of California Press.
Justin Hendrix:
Julian, I'm very pleased to have you on the podcast, and I should say that your book was just recommended to our listeners by a previous guest who was Dr. Alondra Nelson, who flagged it during a conversation we were having as part of a series we're calling The Edge Dialogues. And so I'm pleased to be able to, I think, deliver on her recommendation that folks should pay attention to this book. It was one of the things that she suggested that folks should be reading if they want to follow up on some of the ideas that she had discussed. I want to ask you first just about your research and your areas of intellectual curiosity. I know you look after social, cultural dimensions of information, labor, infrastructure, and the development of AI. What do you think of as your vocation? What is it you're studying?
Julián Posada:
So I see myself as both a sociologist and information scientist, which have my both graduate degrees. I look into society and its interplay with technology and especially with labor. So I started doing this research actually on the gig economy around 10 years ago when I was an intern at a tech company in France with media. And I initially was interested by streaming. I was at the gaming division of the web company and I had a lot of interactions with gamers and streamers. And interestingly, my very first thesis for my master's degree in sociology was about a streamer who would stream the National Assembly debates, not games, the National Assembly debates on Twitch. And the first project was about how these groups of teenagers would build knowledge, political knowledge together through streaming. But then from Twitch, I became very interested in the gig economy. It was a time where delivery apps like Uber Eats was beginning.
I was in France at the time and I was recruited by the Laboratory of Computer Science at the French National Center for Scientific Research to do a project on restaurants and how apps would change the work of restaurateurs in France. And then I decided that I liked research, that I wanted to be a researcher myself and to do a PhD. And my initial idea was to continue doing work on the economy, but that was around 2018 when people started talking about the ethics of artificial intelligence, the issues that AI could pose to society. And this is when the first wave I would say of AI ethics emerged where everyone was talking about principles, principles of AI, what to do with AI when AI was deployed and so on. And as a labor scholar, I was actually very upset to find that rarely work and labor was even mentioned in a lot of these principles coming from institutions like universities, international organizations and companies themselves.
And when it was mentioned, labor, it was always about the future of work, about something that would come in the future and so on. But from colleagues like my mentor, Antonio Casilli or Mary L. Gray who were the pioneers and also Lilly Irani, pioneers on data work or time micro work, the labor of people developing data sets for the AI industry and platforms, large platforms at the time like Mechanical Turk and Appen, I realized that no one was talking about this thing, that the perspectives of the workers were not being mentioned. And that was actually one of my first publications titled The Future Work is Here, precisely to say all of these discussions about how AI will impact workers and so on, it's already happening because AI depends on workers. And as a Latin American, there was really very few papers or basically no research being done at the time about Latin American case and about workers from Latin America.
So that was the beginnings of this book around eight to 10 years ago.
Justin Hendrix:
And you grew up in Colombia?
Julián Posada:
Yes. I'm from Cali, Colombia. My dad was a factory worker as well. So a lot of the ideas from he was a member of a cooperative and all of these ideas from worker struggles come from his work as a factory worker. And yes, I grew up in Cali, speaking Spanish. I moved to France for my studies later on when I was 18. So yes, I'm from Colombia, neighboring Venezuela, which was a case study for the book.
Justin Hendrix:
Absolutely. And I wanted to kind of come to Venezuela next, ask you how Venezuela ended up being the sort of essential focus of the book.
Julián Posada:
Yeah. So it was 2019. We won a grant between my advisor, Alessandro Delfanti, Antonio Casilli and Paola Tubaro to expand this network that we had, European Network on Digital Labor and make it international. So we organized the first international conferences. And it was in the 2018 conference in Paris, I remember well, where one of our colleagues, Florian Schmidt from the University of Drexel at the time, he wrote up web traffic analysis and he was saying, "I used the Alexa RANCAP," which existed at the time. And look, the things that we know about online labor, because there was a big study by Mark Graham from Oxford about online labor and really, they were showing really this North South divide on labor. And for them it was just online labor, right? So including freelancers, designers, programmers and so on. And they were saying that the research found that yes, it was a north south divide that many countries, usually with robust outsourcing industries like India and the Philippines, Bangladesh were represented.
And Florian Schmidt came saying, "This is the same case for data work, but look at Venezuela. It was 2019. It is a very important... There's a lot of traffic coming from Venezuela and we don't know why." And I, as a PhD student, like I said, I was looking for the project to do for my dissertation and I told myself, "This is it. Very few people are doing research on data work as a whole, but no one almost is doing research about the Latin American case. I speak Spanish, I can interview folks, so let's do it." And this is when in 2019 I started developing this project for my dissertation. I actually was very lucky to get funding for field work in Latin America. And then the COVID-19 pandemic hit. So I had to do everything online. And this data collection process then starting in 2020, 2021, completely remote because of the pandemic restrictions and also made actually this case study not only about Venezuela, but also about the pandemic times and how it affected data work as well.
And actually initially wanted to do Latin America as a region, but because the vast amount of workers and traffic from platforms was coming from Venezuela, it ended up being a project about Venezuela. I did interview people in Columbia as well, but the vast majority of workers and interviewees for this project were in Venezuela.

August 13, 2020, Venezuela, Caracas: A woman in a face mask looks at her mobile phone when she passes a mural in Caracas. (Photo by Pedro Rances Mattey/picture-alliance/dpa/AP Images)
Justin Hendrix:
So I think when folks typically think about data work or think about business process outsourcing work, we think of the Philippines, we think of India, we think of maybe Kenya these days. I was not aware before reading this that Venezuela ranks so highly amongst the countries that do data production work.
Julián Posada:
Yes, it was very unique and the book explores, it was a combination of factors. One of them is the crisis in Venezuela. So hyperinflation, highest levels of inflation in the world. Some of the interviewees were telling me we need to check the exchange rate every hour because the money that you have at 9 a.m. could disappear by the end of the day. So you need to really know the exchange rate every hour. And people started using the US dollar as a black market currency. So you have the hyperinflation, you have the COVID pandemic as well, so high levels of unemployment, a lot of people working on the informal economy as well. So that's the first factor. And the fact that yes, many people lost their jobs, but also if they kept their jobs, if they were paying in the local currency, the Bolivar, then their earnings were wiped out by the inflation.
So you have the economic crisis and then you have the infrastructure of the country. People having subsidized electricity and internet, although unreliable, but also many, many people were using computers that were built in Venezuela by the government before the crisis that were given to school children for free. The 2001 child, one laptop program, the Venezuelan version of that was the Canaima computer. And so many people were working with infrastructure that was built by the country before the crisis. And this is why one of my colleagues at Yale was asking me, "Yes, there's many countries in crisis with inflation and so on. Why Venezuela?" And I was telling her the second reason is precisely the infrastructure. You need computers, electricity, and internet for people to be able to work. And then the other reason, actually, the reason for the title of the book is the platform.
The platform is an infrastructure for work. So for a BPO or any previous way of resource extraction in Latin America, you need to build that infrastructure. You need to build ports, you need to build mines, railroads for BPOs, you need to build the facilities, staff the facilities with computers, the security process and so on. For the platform, you really need just accounts and computers, electricity and the internet. You source that to the worker like the gig economy has been done a decade prior. And for this particular type of work, so late 2010s, early 2020s, a lot of the investments on AI were poured towards computer vision. This was a time where the industry was developing facial recognition technologies, self-driving vehicles. A lot of it was image content. And so skills, for example, required for English language that came more for LLMs later on that were more required for that were not necessary at the time.
And so workers could tag images. It was also the time of supervised learning where this is machine learning technique in which they need a lot of label data. So workers were asked to tag roads, tag images, objects in warehouses, for example. Those are the types of tasks that workers were doing at the time. And a lot of Venezuelans were capable of doing it. They had, again, the infrastructure and they were pushed by the economic crisis towards these platforms to do this work. So it was all a combination of factors really that made Venezuela one of really the top players in data work at the time during the pandemic.
Justin Hendrix:
The second chapter of the book opens with Roberto, who is in a neighborhood outside Caracas. He is 17, high school, you say, when he started working for online platforms. You say he was one of the first and one of the last people you interviewed for the book. Why did you bookend it with him?
Julián Posada:
It was because first he was I think in his early 20s when I interviewed him for the first time and he had been working among the interviewees for the book. He was one of those with most experience actually, because like you say, he began when he was in high school and he had been doing it for a very long time. So his story is that he was a university. So he started doing high school, he then went to university, but then the pandemic hit and then he had to go back to the platforms during the pandemic, which was the first time I interviewed him. And he also was very keen into this research. He was really interested about what I was doing. Again, he was an engineer student, so he was working towards his bachelor's degree. So that's one of the reasons I kept contact with him as well.
He could give me a new perspective, I would say, both as an engineering student, but also as a data worker with a lot of years of experience doing this job. So I actually wanted to ask him at the end when I interviewed him again, he was still doing it and like many, many of the interviewees, he said no, simply the situation changed when the Russian invasion of Ukraine and also the launch of LLMs. So the situation in Venezuela improved and the industry moved away from computer vision towards large language models. So a lot of those jobs then appeared in, like you said, Kenya, the Philippines and India later on. And many Venezuelans I interviewed last year for the book, they said the situation just became better and we could go back to our normal jobs or just find employment elsewhere. The situation of the pandemic, which was very dire, had disappeared and also speaks as well to what I call in the book, the extractivist cycle, this way of platforms so malleable that companies can just find work in different countries and just a few months in between.
So that's why I ended up with Roberto. His perspective was very unique for the project.
Justin Hendrix:
So you do attempt to kind of give us a picture of what this industry looks like, how many people are employed, the demographics of it generally. Anything more you'd say about that, about what we know effectively of how many people are data workers on the planet at the moment?
Julián Posada:
Yes, it's very tough. So it's an invisible population and this is why the book required a mixed methods research. I was very lucky that Jeffrey Boase, a sociologist at University of Toronto was one of my committee members. He's a methodologist, very strictly with social research methods. And because it's an invisible population, we don't know how many data workers are there. One of the findings of the book is not even the companies that employ them know how many data workers are. And the reason I'm saying this is because I interviewed the family that the book begins with the story of a family. This family worked under one account, but there was the mom, the dad and the kids as well working all under the same account. And then through interviews with other workers, I found that some of them could have 10, 15 accounts that would just rotate between them to get more money from the platform.
So in the eyes of the platform, they could be just one worker, one account, but could be a family behind them or could be just one person having several. So not even the companies know how many people are working for them. In many cases, they don't know that children work for the platforms. I know from Roberto actually mentioning some minors using their grandparents' faces and names to authenticate the accounts, but in the reality was those children working for the accounts. So really the outsourcing model makes it that so that no one knows who these people are completely. And so I had to use many methods to try to find more about the population. So web traffic analysis like Florian Schmidt back in 2019, trying to estimate the flows of connections between platforms and users and clients to estimate really the global division of work, but also service to workers, surveys to workers and interviews and analysis of instructions, really any data point that I could get my hands on really to understand these labor arrangements better because we were dealing here again with an invisible population that is very hard to estimate.
Justin Hendrix:
There's so many concepts in the book. One that I want to ask you about is the concept of platform workers and what you call epistemic dominance. Walk me through what that means.
Julián Posada:
Yeah. So one of the main claims of the book is to say that there could be extraction of value from the production of data and the exploitation of labor of these workers, but it's not one directional. It means that there is also another way in which workers receive instructions and are made to think in certain ways to then encode that knowledge into the data. I give two examples from the book. One of them is for a facial recognition system that we're asking the workers to categorize people according to race, which today the industry doesn't do that anymore. It's one of those terrible examples that we can discuss from like 10 years ago only. And interestingly enough is that the taxonomy that the workers were given was from the US census, which classify people according to race and ethnicity. But for the workers in Venezuela, that taxonomy designed for the US census did not make sense because racial classification and ethnicity is thought very differently across the Americas and especially in a Latin American country like Venezuela.
So workers were having a lot of struggle with that particular task. And another example that I give in the book is about hate speech. So they were training a model to do content moderation of language and they were asked to label data according to just a binary classification here, if it's hateful or not. The problem is that classification was made following the US law again and First Amendment, which is very, in the world really the freedom of speech in the US is very extensive. I always give my students the example of flying a Nazi flag is allowed here because of freedom of speech grounds, but it's criminal, a criminal offense in Germany, for example. And in the particular case, I bring in the book, a worker that I interview was asked to label a comment made by someone who would come up with something like, "I want to kick all the Latinas out of my country," which the worker found as a Latin American, the worker found very offensive and wanted to label it as such, but then it was rejected.
The platform was arguing that no, it did not counter these ideas of hate speech that the client had given the platform. So this is just a couple examples to show how a lot of the data labeling comes from the top to the bottom. This is based also on research that I did with Milagros Miceli. And because the workers are in a precarious position, they don't have a contract, they're users of a platform, their entire family's in a crisis situation, so the families in many cases depend on their income. They cannot just disagree with what the platform the client is saying. So they always have to obey these instructions that they receive from the platform. And it's sort of this idea that they are being imposed these ways of seeing the world, imposed this ways of knowledge. A lot of the exercise of the book is to frame platform extractivism not as something new, but as a continuation of older extractive legacies that are also tied with older capitalistic and colonialist systems.
And I begin actually that chapter, the epistemic chapter with this example of the Spanish conquistadors coming to the Americas to destroy, not only to extract value from the land, but also to impose their religion, the ways of seeing the world. And here with the instructions, what I saw was this same type of imposition and hence why I call it epistemic dominance, a way to dominate in this case, the minds in a way through instructions and through the entire process of data production that these platforms enable. And this is tied to other forms of other concepts in the literature like epistemic violence from Spivak and so on that also talk about this way of imposing knowledge. Again, the point here is to say that there is extraction of value from data and work, but it's also this imposition of knowledge that is made to workers in this case.
Justin Hendrix:
What did you find out about what happens to a worker who disagrees with the label that they're being made to apply?
Julián Posada:
So they are usually, it's called kicked out of the task, meaning you cannot continue the task. Some cases workers are given warnings and depends on the platform. Other cases you just stop working for the task. Some more extreme cases, they are what they call banned from the platform, which is basically being fired from it. The problem is when that occurs, it occurred to actually Roberto, the worker that you mentioned, he has dark skin. He doesn't identify himself as black, but he has dark skin. And he said that when they implemented a facial recognition technology to authenticate his account, he was having trouble because the facial recognition system did not identify him. I know from research by Timnit Gebru, Joy Buolamwini, for example, that this initial facial recognition models were very inaccurate with people with darker skin. And because he could not authenticate the account, he could not access it, so he was basically unable to work.
And the platform basically did not pay him for the work that he did. One of the things is that many of these platforms would pay you every week. And if you, for example, were banned or kicked out of a task or if you lose access to your account, you would not get paid. But also there were some platforms Hive, especially you would have a threshold of, I think it's $15 per week, 10 to $15, and if you would not reach that threshold, then the platform would not pay you that week. So it was also a way for, one, force workers to work many, many hours, but also what is considered as wage theft. If you don't reach the thresholds, you'll not get paid for the work that you have already done. And this was not uncommon to hear from the workers that I interviewed for the project.
Justin Hendrix:
I want to ask you about examples of worker collective power. This is another kind of focus in the book, how workers are able to organize, able to kind of advance their own interests. What examples do you see that are emerging?
Julián Posada:
Yeah, so there's some unions that have emerged in data work globally, especially the data workers labor union in Africa, the content moderations union also in Africa. In Venezuela at the time, many workers saw this because there was a crisis in the pandemic. They saw it as a temporary measure, and yet they found many, many ways to find other workers to discuss the tasks. They used a lot of social media platforms for this like Discord, WhatsApp, and Telegram. The most interesting cases I found were chats, like group chats on Telegram and WhatsApp, they were closed, meaning only a few members could access them. And for example, if you wanted to be a member of these groups, for some, for example, you have to show your ID and pay a membership fee. So they were really organized groups in that case. And then they would provide a lot of benefits for workers.
One of the main benefits was brokers. Workers were not paid in US dollars because of sanctions and other reasons, but they were paid through digital currencies. And one of them was, for example, through a Mexican e-wallet platform called RT... What's the name of this platform? I'm forgetting the name. The currency was AirTM. I forget the name of the platform, but basically workers needed then a broker who would then buy or sell the digital currency and locally pay in cash to the worker. And there was a lot of scammers around precisely with scamming through this transaction. So these groups would also allow workers to then find trusted brokers that they could access then to exchange their currency. But also these groups were made for workers to exchange guides, to find exploits on the platforms like bugs or errors in annotations so they could get a lot of money by doing less work in this case and voicing, of course, a lot of complaints, concerns, even ethical concerns about the tasks they were doing and even a form of insurance as well.
So they used the money not only to pay for the administrators, the moderators, but also to pay for tools like bots that were alerted and where tasks were available or even insurance in some cases in which, for example, a worker would lose money from scams or other instances. So these were really organized groups that emerged for workers. I imagine a lot of them stopped working again when the situation improved in Venezuela and many of these groups maybe disbanded, but they really were there and workers really organized in a way that would facilitate their work at the time during the pandemic.
Justin Hendrix:
This kind of thing's always fascinating to me. We had a segment on the podcast a couple years ago where the audio reporter, Rebecca Rand, spoke to Dr. Luis Lozano Paredes about a community of drivers in Colombia who were using WhatsApp, I believe, to hack together a way to effectively game Uber and to try to preserve some of their power in the Uber ecosystem. So I don't know, that stuff's really, really interesting to see the creativity effectively of folks trying to preserve their ideas.
Julián Posada:
There is an example from Jamie Woodcock from London with Deliveroo, which is a delivery app there, and workers trying to recruit other workers were ordering food to the picket line. And that way they would feed the picket line, but also they would bring new workers to the picket line and try to convince them to go and strike with them, which is a very interesting case again of using the platform against the platform, I guess, in this case.
Justin Hendrix:
So you talk a lot about kind of support structures and the way that these individuals who are doing this work band together, but also the importance of their families, their neighbors. There's a kind of, I suppose, in it together kind of mentality about this. You already mentioned that sometimes this work is shared in families, perhaps without the knowledge of the platforms. I don't know what more could you make of that?
Julián Posada:
Yeah, so actually one of the last additions to the book, and I had trouble finding that quote again and be able to cite it. So I play video games a lot and one of them is Cyberpunk and the character in the DLC randomly says, "Worse of you are, more you get to stick together." And I had to find that quote again, it was pretty hard, but that really tells you about the importance of your neighbors, your family, your friends in this situation. So to put things into perspective, so again, Venezuela has a crisis, has a pandemic. What it also means is that the state, which was able to provide many services from hospitals to proper education and other things was not there anymore to provide even basic things like picking up your garbage, right? So you have the state receding in that case because of the crises that were happening, and because you are an independent contractor for a platform or just an user of a data work platform, you also don't have your employer to provide any social support for you as a worker.
So what in all these cases people ended up doing is to rely more and more on their personal network. So again, family, both locally and abroad, neighbors, anyone within the community that could provide any type of social support. And that could mean many things. It could mean, for example, if you have kids, then mom or aunt could take care of the kids while you're in front of the computer while doing data work. It could mean if you lose your electricity or if you lose your internet, you could go work from your neighbor's living room or your relative across the city's living room, for example. It could mean as well for things like water or waste management, like I said with garbage, that the community itself were organized then to provide with these basic services. So this is when I try to bring this notion of territory or territorial from Latin American struggles, especially indigenous and campesino struggles, to understand this wave of relationships, not only between people, but also with the land, but how important it is, all these networks and relations that have occurred among many, many entities.
And to bring up the claim from the book really that it is not a story about individual workers, but really about their entire communities, their entire territories that is at the core of this extractivism that I mentioned in the book. And I would say a third reason then I mentioned the crisis, I mentioned the infrastructure, but the territory is really the third part, the third piece of the puzzle here that explains why data work was so prominent in Venezuela and how can companies find thousands and thousands of workers to produce data for the AI industry for very cheap in the eyes of the platform.
Justin Hendrix:
So I want to ask you about this idea of the false promise of AI. This is kind of where the book lands us. And you talk about this idea of the freedom of work and the freedom to work. Tell me about this false promise. I mean, I suppose everyone, you and I both live in New York City. I get on the subway in the morning and every promise I see in the advertisements is either about how AI is going to make my work life so much easier or otherwise kind of enable me or it's going to somehow address some other urgent need that I have. But why is this a false promise? Why do you see a false promise here?
Julián Posada:
Yeah. Now that I mentioned New York City, actually when I explained the promises, one of my slides is a photo I took from the New York City subway of one of the ads precisely. So the two promises, freedom from work and of work. This one is to gig workers like the data workers, is this idea that you can work from anywhere, the amount of hours that you would like, you're your own boss, a lot of flexibility, which is actually a lot of workers, and this is something I totally agree with my colleague, Payal Arora, and something that I stress a lot is that from all the workers that I interviewed actually, they have a good look or a positive view of the platforms because really for a time of crisis, these platforms allow them to earn some income at a time they really, really needed it.
So there's this positive look that workers have on the platforms. However, they can work at any time, but only when tasks are available. They can be their own boss on paper, but really it's the platform that controls the algorithm, that ranks them, tells them how much they earned, gives them specific tasks, and also can fire them at will, and they don't have to explain anything if they get fired. There is a lot of flexibility which workers truly value, but the power really is in the hands of the platform that decides really from wages to availability of work and who gets to work when and how. So really it's a false promise in that sense. You get to work, but really only under the terms that the platform decides. And the other promise is freedom from work, which is for all of us here, that AI can increase our productivity and make us work less hours in the future where machines can do all of the menial tasks for us, then we can have more time for leisure and other things.
But really what we're seeing historically from labor economics is that even though productivity has risen and unemployment actually has remained stable, meaning we actually do a lot more with machines, the wage gap is larger and larger. We're in a K-shaped economy, some economists mentioned, there's a lot of wealth being in the hands of a few and a lot of people don't see that wealth trickling down. And in terms of labor, we see a lot of outsourcing continuing and the gig economy actually impacting more and more jobs. So one of the platforms that I studied in the book, Remotasks, its parent company is Scale AI, which is backed by Meta, and actually the CEO of Scale AI is now part of the Meta AI team. And they released a new platform a few years ago called Outlier. Outlier is for LLMs and what they want is to recruit what they call expert labor.
So it means that they're recruiting people with knowledge of specific languages, but also knowledge of programming, math, law, and medicine. One of the recruitment ads that I looked at just a few weeks ago, it was to hire a cardiologist in New York City for $180 an hour to look at x-rays and other type of medical imagery to train a model to identify illnesses and other types of issues related to then x-rays and other types of images. What I found very interesting is even though they're looking for someone with an MD, a medical degree, paying them almost $200 an hour, the terms and conditions of that employment are exactly the same as the terms and conditions of the Venezuelan workers during the pandemic. They really just copied and pasted the terms of service from Remotasks into Outlier. So what we're seeing here is the gig economy just getting in other sectors, even to types of employment that we would think save from AI, like medicine or law, which are really high paid jobs, even in a country like the United States.
So we're seeing this transformation of labor emerging, the gig economy again, going to new sectors, outsourcing, increasing, and the productivity increasing, but not the wages, not the wealth of a lot of people. It's not really trickling down. So this is why I say there's also this other false promise. We don't see yet the fruits of this freedom from work that many believe AI would bring to society.
Justin Hendrix:
The idea that we'll all become gig workers would seem to fly directly in the face of the abundance notion that people are selling with AI.
Julián Posada:
Correct.
Justin Hendrix:
So let me ask you about policy where you see opportunity, where you see progress being made. Where do you think the policy environment appears to be supporting collective action by workers?
Julián Posada:
I mean, there's a lot of emergence of, like I said, unions and other forms of worker organizing in many places around the world, especially in Africa. Many governments like the European Union are starting to think about data work. Like I said, back to the beginning of the podcast, this idea that 10 years ago, a lot of the documents about principles of AI coming from places in the global north were barely mentioning workers in the global south. A lot of agencies, governments, international institutions are actually starting to talk about data work and the importance of workers in the global south for AI development. But one of the things that one of my colleagues, Mark Graham and the team at the time, thinking again about online work mentioned is that because the flexibility platforms allow for employers to just move from one country to another, and it happened actually.
There was this company called Sama that just shut down. There was a lawsuit against them and their main client was Meta. So there was a judge ruling against Sama continuing some of the work that we're doing because of egregious labor conditions in Kenya. And they may not just hire another subcontractor in West Africa in Uganda to continue this work. So back to what my colleagues were saying, thinking about online labor is that the platform allows for a lot of flexibility for companies to just move to another country where the laws and the government of the country are more friendly towards the platform and the clients, which is again, speaking about the extractivity cycle from the book, how easy it is to just find workers in other countries because the platform as an infrastructure for work allows for this. So actually we should hope really that a lot of the regulation for this comes from the countries to which that labor goes to, the fruits of that labor goes to.
So that's the United States. China has their own, they're more national in terms of their data work. It's mainly workers from China doing data work for China. But for here, the United States, most of these workers come from abroad. So we would hope that, especially the new Congress that was elected, would do something about this. The previous administration through USAID had some policies in place precisely for data work to ensure decent working conditions across the entire AI data pipeline. So I'm just hoping that the new Congress that was elected that would be more open towards regulation of AI than would address some of the issues that the book discusses. And even though the book, actually the last full chapter of the book talks about what I call top-down approaches like regulation, but always let's remember that worker agency, worker organizing is very important as well.
So those bottom-up approaches, what I call in the book, are also very important to ensure that, again, decent working conditions are for everyone in the AI data pipeline.
Justin Hendrix:
Well, we will see if any of those types of reforms are carried out either in this Congress or elsewhere around the world. We're going to be paying attention to your work, Julian. Platform Extractivism: Data Work and the People Powering Artificial Intelligence is out from the University of California Press. And I hope listeners will go and pick it up. Thank you so much for joining me.
Julián Posada:
Thank you, Justin, for having me here.
Authors

