Enshittification With Lina Khan at the Brooklyn Public Library  ## Metadata - Author: [[Podcast – Cory Doctorow's craphound.com]] - Full Title: Enshittification With Lina Khan at the Brooklyn Public Library - Category: #podcasts - URL: https://share.snipd.com/episode/3e820f5e-ce94-4778-bc90-7ad4ed473be6 ## Document Notes ## Highlights - **Policy, Not Just Greed, Drives Decline** - Inshittification is a systematic business strategy where platforms lock users, then worsen services to extract more profit from captive customers and business partners. - Cory argues the root cause is policy choices that removed market and regulatory discipline, not merely greed or consumer mistakes. Transcript: Cory Doctorow So I think that a lot of people were initially attracted to inshittification's descriptive component, right? This pattern we see with platforms where they're good to their end users. They find a way to lock those end users in. And then because they know it's hard for those end users to leave, they can make things worse for those end users and make things better for business customers who get tempted in. And then they can make things worse for those business customers too, because they become dependent on those users who are locked in and they can harvest all the value for themselves. Now, there have been firms that have worsened their offerings in the past, but historically what's happened is they then died, right? We switched to someone else. They went into business. It's not considered to be a viable strategy to be someone who makes a product everyone hates. And yet what we have now are these kind of zombie firms that are like so past their sell-by date, you know, long past the point where someone should have like double-tapped them and stuffed Them in a shallow grave, but we can't seem to quit them. And the most, I think, compelling part of the incentivification thesis is really the explanation for why this is happening now and why it's happening to so many of those firms. And that's a policy explanation. ([Time 0:03:48](https://share.snipd.com/snip/fc8aeeab-5ad5-4c96-8df1-03c1dc66ecf2)) - **Concentration Enables Surveillance Economies** - Lack of competition and regulatory capture erode firms' incentives to protect user privacy and service quality. - Cory links market concentration to weak privacy laws and explains why incumbents block meaningful regulation. Transcript: Cory Doctorow So, you know, part of the idea of inshittification is tracking the collapse of discipline in firms, the things that punish them for doing bad things. So, you know, my thesis is that if you don't have competition, then you're not worried about your customers defecting. But also if you don't have competition, if you're cartel, it's easier for you to capture your regulators. You know, if there's a hundred companies in a sector, it's really hard for them to agree on what line they're going to feed during a proceeding. And they are competing with each other, they don't have as much money to push that line when there's like five of them, they can really agree on a position and privacy is a really good example Of this. So our last consumer privacy law federally in this country was passed in 1988. And it's a law that bans video store clerks from disclosing your VHS rentals, right? And like, it's not an accident that we haven't updated privacy laws since Die Hard was in theaters, right? It's because there's a cartel of extremely powerful firms that make a lot of money from it. Now, one of the other aspects of the initiatification hypothesis is that it's not just end users who are getting screwed. It's the businesses, too. And so, you know, Google and Facebook and these other firms, they have a story about what they can do with our data that doesn't bear up to scrutiny. ([Time 0:10:16](https://share.snipd.com/snip/0359fd36-fea7-48cc-9509-5b28c6526592)) - **Ad Surveillance Is Overclaimed** - Surveillance economics overstates ad efficacy; ad performance often fails to justify massive data collection. - Cory cites Procter & Gamble cutting ad spend with no sales loss as evidence of advertising wastage. Transcript: Cory Doctorow Right. You know, Mark Zuckerberg basically claims that he's an evil dopamine hacking wizard who has figured out what like Rasputin and Mesmer and MKUltra couldn't figure out, which is how To make a mind control ray by taking your data. And he uses that to sell ads, right? Well, I think that, you know, the evidence for this is quite poor. In fact, when you actually look at ad performance, it's not great. And, you know, not only that, but he was so lackadaisical about policing ad fraud that a lot of the ads that he was supposedly manipulating us with just were never shown to anyone. Procter& Gamble in 2017 zeroed out a $200 million per year annual advertising spend and saw a $0 drop in sales. Because to a first approximation, $200 million in advertising was just disappearing down the ad fraud hole because it was cheaper not to police ad fraud than it was to police it. When you have a monopoly, you can get a lot of money from people for selling them defective products. But the fact that our private data is being sucked up as part of this kabuki about the ability to change our minds from it does in fact open up all kinds of vulnerabilities. ([Time 0:11:28](https://share.snipd.com/snip/648e2373-f955-4441-8422-6e8d20e5a781)) - **Surveillance Harms Build Broad Coalitions** - Commercial surveillance creates concrete harms: targeted exploitation, price and wage discrimination, and facilitation of fraud and abuse. - Lina and Cory argue these harms unite diverse coalitions for privacy reform beyond abstract privacy debates. Transcript: Cory Doctorow So one was a speaker from the DOD who talked about the fact that ad brokers or data brokers will sell you a category to target ads to that is people on army bases with problem gambling, right? And then there was someone from the AARP, which, you know, frankly, if I were a politician, I would be terrified to cross that person because they vote like crazy. And she said, oh, yeah. And another category that you can target with an ad broker is people with cognitive impairment, with age-related cognitive impairment. And so you can see how you don't have to have a mind control rate to really hurt people with the surveillance data. We have breaches, of course, that harm people. But then you have things like price discrimination with surveillance data. And we all worry about the day that Walmart puts electronic shelf tags on the stores and then doubles the price of an umbrella every time it starts raining. But one of the most grotesque forms of price discrimination is wage discrimination. When a firm pays you less than someone else for doing the same job, they're saying you are worth less than they are. Just like when a firm charges you more than someone else for the same product, they're saying your dollars are worth less than theirs. And so in this country, hospitals preferentially hire nurses as contractors and not as staff, because they're doing union avoidance. And historically, you would contract with nurses through a local staffing agency. And there'd be one or two or three of those, depending on the size of the market. Those have all been wiped out by four apps that all call themselves Uber for nurses, right, in their marketing materials. And before a nurse is offered a shift, the app goes to an unregulated data brokerage and says, how much credit card debt is this nurse carrying? And the more credit card debt the nurse is carrying, the lower the wage they're offered on the grounds that they're inferring economic precarity and desperation. Now, obviously, this is a very bad deal for a nurse. It's a very bad deal for society. But it's a bad deal for the person getting a catheter inserted by a nurse who had to work till midnight the night before driving Uber and couldn't afford breakfast that morning because It was breakfast or rent, right? This is just a bad deal all around. There are so many harms that start with this great void in our privacy rules. ([Time 0:12:56](https://share.snipd.com/snip/4672ff7d-b3f0-4ade-8aeb-ccbb5edb95e4)) - **Labor Was A Key Check That Eroded** - Tech workers once constrained bad product choices by holding employers accountable through scarce talent and mission-driven loyalty. - Massive layoffs and weakened labor power removed that internal check, enabling inshittification. Transcript: Cory Doctorow Yeah. So, you know, like I, in the book, I break down four sources of discipline against inshidification that firms have historically faced and talk about the factors that eroded. So we talked a little about competition and regulation. The third one is labor. And the tech workforce, it's a historically unique workforce in that it was historically very powerful, but it was not unionized in any great amount. The union density in tech could barely be detected with a microscope. But the thing is, tech workers were in short supply and they were extraordinarily productive. There's a National Bureau of Economic Research working paper that estimates that the average tech worker in Silicon Valley contributed a million dollars to their employer's annual Bottom line. This is why you got gourmet cafeterias and free kombucha and massages and the surgeon who'd freeze your eggs so you could work through your fertile years, right? It wasn't because they were sentimental about tech workers, right? It was because they couldn't replace them if they left. And people were trying to hire them because they wanted to get the million dollars per worker that you could get if you could bring a worker in. And it is a great irony and paradox of labor power that workers who are powerful and therefore treated well because of the historically contingent circumstances don't think they need A union. And really, the time to form a union is when you're powerful, right? When you lose your power, that's a much harder time to unionize as tech workers are discovering. And you know, tech workers aren't the only workers who work for tech companies. Tech bosses have shown us how they treat workers they're not afraid of. Jeff Bezos' warehouse workers are injured at three times the national average, seriously injured. Jeff Bezos' delivery drivers have to pee in bottles and they're monitored by cameras that ding them if they open their mouth while driving because you're not allowed to sing on the job, Right? Jeff Bezos's programmers get to show up for work with pink mohawks and facial piercings and black t-shirts that say things their bosses don't understand but feel a little uncomfortable About. And it's not because Jeff Bezos likes programmers, right? The minute Jeff Bezos can put a piss jar next to your workstation, he will, right? For a man with penis shaped rockets, he really hates our kidneys. And the way that tech bosses motivated tech workers was appealing to their sense of mission, and it's not a hard reach Because so many people's path into tech is discovering their people, The words for the things they feel, the milieu that they belong in, by joining online communities, by getting online, by discovering a much wider world out there beyond the confines Of their home using technology. And they want to bring that to other people. And it makes them, I call it, Tron-pilled, right? They want to fight for the user. And when your boss says, oh, we're like on a holy mission, right? We're going to organize the world's information and make it accessible. Or we're going to bring the world closer together and make it more connected. You believe it, right? And you want to do it. And you do get to do it. Because when your boss says, well, it's time to unshittify that thing you missed your mother's funeral to ship on time because you believed in it so much, you can tell your boss to go screw Himself. Because there's a guy across the road who gives you a job that afternoon, and your boss can't hire anyone to do your job. And so you get to hold the line. You get to be legitimately trumped. But supply catches up with demand. We have had half a million tech layoffs in this country in the last three years. ([Time 0:20:06](https://share.snipd.com/snip/14de4d70-bb13-4d50-9155-926e149ab8b1)) - **Products Can Be Altered After Sale** - Digital products can be tethered and changed post-sale, unlike physical goods, enabling firms to degrade products or convert features into paid add-ons. - Legal rules like the DMCA's anti-circumvention provisions amplified manufacturers' control after sale. Transcript: Cory Doctorow Yeah, so I'm a pretend computer scientist. I have an honorary degree in computer science. And I dropped out of an actual computer science program. But a computer scientist will tell you that we only know how to make one kind of computer. It's called the Turing Complete Universal von Neumann machine, which is a lot of jargon, but what it means is like every computer is an engine capable of computing all valid programs. And practically what that means is if someone puts like a 10-foot pile of shit in your program, someone else can make an 11-foot ladder that lets you go straight over it, right? If you lock third-party ink out of the printer, someone can unlock the printer. If you lock independent mechanics out of the car, someone can unlock the car to do it. If you steal people's private data with an app, someone can make a privacy blocker for the app. And, you know, at the start of this talk, I said, well, some of the policy predecessors took decisions that at the time they were warned would have this initiatory effect downstream, Did it anyway, never faced any consequences. And one of these is a guy called Bruce Lehman, still alive, who was Bill Clinton's IPs are. And he had this idea for a law that would make it illegal to reverse engineer and modify any product that was designed not to be reverse engineered. So in other words, if you just had like a box around some code that said, don't change this code, looking inside the box or changing things inside the box would become illegal, even if You're doing it for a legal purpose. He proposed this to Al Gore when Al Gore was doing the information superhighway hearings, the national information infrastructure hearings. And to Gore's eternal credit, he said, go away, that's a terrible idea. And so Bruce Lehman went to Geneva, to the World Intellectual Property Organization, this UN specialized agency that does IP law, and he got them to pass these two treaties, the WIPO Copyright Treaty, the WIPO Performers and Phonograms Treaty, the Internet Treaties. He went back, he got Congress to pass a law called the Digital Millennium Copyright Act to adopt the terms of this treaty. Later, he said, I did this quote to do an end run around Congress, right, to turn what had been rejected democratically into an obligation. And under Section 12.1 of the DMCA, it is a felony to modify a device that is designed not to be modified, to bypass an access control, and not a little felony, a five-year prison sentence, $500,000 fine for a first offense. And what this means practically is that firms who design a product so that you can't change it can themselves change it as much as they want, and you can't change it back. So digital products, you know, they are unique in that they can be tethered to their manufacturer after sale. If a manufacturer later regrets having added five features to the product that maybe they could have sold you as add-ons later, they can take three of them away and make them add-ons. And indeed, because these products generate so much telemetry, they can look at which features you use the most and take those away and charge you. And because they can do it on an individuated basis, they can do it per user and change the offers on a per user basis. And you as the end user can't go in and change it back. ([Time 0:26:23](https://share.snipd.com/snip/b1f93b4e-ffaf-4d92-90bf-55cf6305068c)) - **Garage Openers Forced Into Ad Apps** - Chamberlain bought rivals and remotely disabled an open HomeKit feature, forcing users to use their ad-filled app. - Cory uses this to illustrate how lack of competition lets firms change products against users' interests. Transcript: Cory Doctorow And so to pull this all together with the lack of competition and regulatory capture, it's a company called Chamberlain. They were a garage door opener maker, but they got a bunch of money from private equity and they bought all the other garage door opener makers. Whichever garage door opener you have in your garage, I guess New Yorkers don't have garages, but whichever garage door opener your relatives from the city you moved here from have In their garages, no matter what it says on it, it's a Chamberlain garage door opener. And Chamberlain garage door openers all supported a thing called HomeKit, which is a standard way to interface an app with a garage door opener. So there were lots of different, and other home automation, smart light bulbs and thermostats and whatever. So there were a bunch of different apps you could use to open your garage. So Chamberlain reached into all of those garage door openers and they flipped the switch that turned off HomeKit. And then they said, you have to use our app. And to open your garage, you have to look at seven ads, right? Now, like you can just see how, when there's no discipline on the firm, right? When there's no consequences for doing this, because people aren't going to rip their garage door openers out and put in new ones. And even if they do, which garage door opener are they going to buy? That's not a Chamberlain garage door opener, right? You remember Lily Tomlin used to do these bits on SNL where she would pretend to be an AT&T operator. She'd do these ads for the Bell system. They'd always end with her turning to camera and saying, we don't care. We don't have to. We're the phone company, right? When you have a 90% search market share in your Google, you don't care. You don't have to get the search company. When you have a 90% garage door opener market share, you don't care. You're that garage door company. And when you're digital, you can reach into that product and you can change the rules. I call it the Darth Vader MBA, right? I'm altering the deal. Pray I don't alter it further. And so your products, you have no steady ground to stand on. You can read consumer reports till your eyeballs fall out of your head. You will not know what that product does because you won't know what it does in the future. ([Time 0:30:18](https://share.snipd.com/snip/353c9a64-d5a0-40d5-97ad-03b948d9d8d4)) - **AI Amplifies Incentives To Extract Value** - AI will amplify existing inshittification mechanisms by enabling price and wage titration and replacing workers with lower-quality automation. - Cory predicts large capital burns in AI will create a bubble that could collapse and reshape the industry. Transcript: Cory Doctorow So one thing machine learning is very good at is doing things like titering the wage offer to Uber drivers or nurses down or titering up prices and kind of analyzing usage data and figuring Out how to charge you more or pay you less. So that's, I think, the one obvious way. There's very little about AI in this book. And the reason for that is until the start of the summer, I was just so sick of talking about AI that anyone who mentioned AI at one of my talks owed everyone else a drink. But I eventually reached the point where I had enough complicated nuanced thoughts about it that I pitched my editor on a book about it. And I wrote it over the summer. It's called The Reverse Centaur's Guide to Life After AI. And it'll be out in 2026. And it is a very skeptical book about AI, but I think it is skeptical about AI in ways that other people are not skeptical about it. And so I don't think AI can do your job. I talked about why programmers aren't going to be viably replaced with AI. But I do think that because employers are insatiably horny to replace workers with chatbots, that AI salesmen push on an open door and that an AI salesman can 100% convince your boss To fire you and replace you with an AI that can't do your job. And I think we see this, for example, with customer service already, right? But also in other domains, frighteningly in domains like deciding whether or not you can keep your kids or deciding whether or not you're going to get parole or writing briefs for your Legal defense. I mean, these are really not applications we want to see AI in. And, you know, the title of the book refers to this idea about automation theory of centaurs and reverse centaurs. So a centaur is a person assisted by a machine who dictates the machine's behaviors, a person's head on a machine body. But a reverse centaur is a human who's been conscripted to be a peripheral to a machine, you know, who's been asked to do the final bit of the work that the machine can't do at the pace the Machine would prefer. You know, this is Amazon warehouse workers, right? They're being asked to keep pace with the mechanisms in the warehouse. This is why they're injured at three times the rate of other warehouse workers. They're not just being used by the machine. They're used up by the machine. Now, the great kind of tragedy and terror of all of this is that AI, if it can't do your job, then whatever benefits we get from it are not commensurate with the capital expenditures that We've made for it. Nor will they be sufficiently profitable to incur the operational expenses associated with maintaining these large foundation models. And when you've gone through $600 to $700 billion in capital expenditures as an industry, and your revenue by your own estimates industry-wide is $45 billion a year, and your assets Are depreciating faster than fish, You do not have a lot of time to make your nut back on this stuff. And the only way you're keeping the lights on is by burning other people's money. And people are eventually going to lose patience with having their money set on fire. ([Time 0:36:22](https://share.snipd.com/snip/93598a1e-fc21-4b1c-ae40-0efce8e7ada2)) - **AI Investment Likely Outstrips Sustainable Returns** - The AI boom may be a bubble: massive capital spent on foundation models outstrips likely revenue, risking a damaging crash. - Cory warns the industry burns other people's money and a collapse could be worse than 2008 for investors and employment. Transcript: Cory Doctorow You know, Enron left behind nothing, right? WorldCom, well, you know, I think it's fine. Like, I'm mostly a prison abolitionist. I actually think it's fine the CEO of WorldCom died in prison. He stole a lot of money from a lot of people who couldn't afford it. But AT&T just bought the fiber that WorldCom put in the ground fraudulently in my neighborhood, and now I've got two gigabits symmetrical internet thanks to the WorldCom fraud. I still would prefer that we didn't have the WorldCom fraud, but there is a productive residue left behind. Crypto, all crypto is going to leave behind. It's like a few programmers are better than secure programming in Rust, but mostly terrible Austrian economics and even worse JPEGs. Whereas AI, well, if you were planning for a future of AI, I would ask yourself what you would do if you could buy GPUs at 10 cents on the dollar, hire programmers and skilled applied statisticians Who are looking for work, and have a lot of open source models that have barely been optimized, and that are already doing pretty impressive things on commodity hardware, and probably Can do even more impressive things if we start to look at ways to optimize them to do even cooler stuff. And we're starting to see the leading edge of that. But I think that a lot of the questions we have about AI are really the wrong questions. Obviously, the question like, will the AI wake up and turn us into paperclips is a profoundly stupid question, right? The idea that like if we just keep shoveling words into the word guessing program, it will wake up and become intelligent is like the idea that if we keep breeding horses to run faster And faster, someday a mare will give birth to a locomotive, right? The difference between humans and word guessing programs is not how many words we guess with. And so adding more words to the word guessing program doesn't make it into a person. But by the same token, asking what we're going to do about these foundation models over the long term assumes that we're going to have foundation models over the long term. And maybe there'll be some that are heavily subsidized. Maybe Trump will bail some of them out. He's good at bailing out people who suck up to him. He did put a tip jar on the Resolute desk and put Trump coin in it. But it's a lot of money they're spending. It is baffling how much money they are spending on this. It is truly grotesque how much money they're spending on this, how expensive this is. ([Time 0:40:14](https://share.snipd.com/snip/6e34554d-3524-4a73-a74c-3685080091fe)) - **Restore Interoperability And User Portability** - Favor bright-line interoperability and user rights so third parties can interconnect and users can leave without losing contacts or data. - Cory recommends reducing legal tools that punish reverse engineering and enabling users to control device behavior post-sale. Transcript: Cory Doctorow I think one of the problems with, with bans on self-preferencing and other, other comparable rules is how hard they are to administer, right? So when you go to Google and you search for weather tomorrow, New York city, and it shows you Google's own weather service at the top of the results, rather than pulling something from Someone else, you have to kind of peer into the mind of Sergey Brin and determine whether he thinks that's genuinely the best way to get the weather or whether it's, you know, an act of Naked self-preferencing. And, you know, these guys, they have like a written culture. They have a big distributed organization. So they often put confessions of guilt in writing, you know. So sometimes you can do this. This is kind of the crux of Yelp's story about Google. You know, Mark Zuckerberg, when he's buying Instagram and his CFO said, why do you want to spend a billion dollars on a company with 12 employees? He like sent him an email back saying, well, like people like Instagram better than Facebook. So we'll eliminate a competitor. Right. Like in an era where like the only thing you're not allowed to do in antitrust is say you are making an acquisition to acquire a competitor. This is like, hey, Bob, you know that guy we're going to kill tomorrow, just for avoidance of doubt, it's definitely a murder. And as I write these words, I am totally premeditating it. Now, Obama's DOJ waved that through. So, you know, make of that what you will. But, you know, those remedies can be hard to administer. So I like some brighter lines. Right. And I think that, you know, if we can say like you, maybe you don't have to like adopt a standard, although sometimes it's good to tell companies to adopt a standard, maybe, you know, in Addition to adopting a standard or alongside adopting a standard to allow people to interconnect with your service and for users to leave and not lose contact with the people that matter To them or the digital assets they have on your platform and so on. Maybe we just take away some of the rights that we've given you to invoke the state to punish people who unilaterally alter how your service works. I think that we're like quite afraid of the awesome might of the technologists who work for these companies to repel borders. And I think that the evidence doesn't support it, right? In the book, I tell the story of a startup that made a little app called OG app. An OG app would get you to log into Instagram in a browser. And when Instagram gave back what's called the session key, the thing that tells you that you're logged in, they would kind of steal it, stick it in their app. And then they would go to Instagram and say, tell me everything waiting for Lena on Instagram. And they would throw away all the suggestions. They throw away all the ads. They throw away all the recommendations. They throw away everything that had been surfaced from three months ago because it was clickbait. And they would just show you the things that the people you subscribe to had posted recently in reverse chronological order and would not deliver any telemetry to Instagram about how You're using it. So like Instagram knows when you're scrolling, it knows the angle of your phone from the accelerometer in it. It knows when you stop, it knows what you tap on. All they would send is if you like something would say this post has been liked. And if you commented on it would say this post has been commented on. It was developed by two teenagers who outwitted Meta. And Meta didn't best them in a match of wits. Meta, in one day, they were in the top 10 of both two mobile app stores. Meta just asked those app stores to delist them, and they did. And that was the end of OG app. So in addition to these structural remedies, these conduct remedies, I think what I guess if we were talking about economics, we would call predistribution. The rules of the game that actually determine who can intervene directly on behalf of users. And what this does is it doesn't substitute for smart, committed, shrewd and canny regulators. But what it does do is hedge against venal, stupid, uncaring regulators who are in the tank for the bad guys. When the new guy at the FTC cancels all the pending investigations and enforcement in favor of setting up a snitch line so you can tell which of your co-workers are woke, right? You don't have to wait for the FTC to show up and say, stop doing that Apple, stop doing that Instagram, right? You have a chance at it. ([Time 0:45:01](https://share.snipd.com/snip/628c2313-b03d-4f60-940d-9fa47fac0e17)) - **Trade Policy Shaped Tech Dominance** - Global tensions and U.S. trade pressure created legal constraints that helped entrench tech incumbents' control internationally. - Cory sees an opening as other countries resist U.S. trade coercion and build competitive alternatives. Transcript: Cory Doctorow The thing that I find very exciting about this moment is that, you know, I've worked internationally for EFF. I was their European director. I worked in 31 countries. I used to be on the road 27 days a month. I didn't plug in my fridge because it cost $10 a month to keep my ice cubes frozen. Law and interoperability law, reverse engineering permission. And they would say, yeah, but the U.S. Trade representative has threatened us with tariffs if we do that. Right. And like having fought the U.S. Trade representative at two removes for more than half my life now and understanding like how good they were at making other countries subordinate their own interests to American Industrial interests. It is quite remarkable to me that Trump, as part of his America First initiative, has neutered the U.S. Trade representative by taking their deterrent and realizing it. When someone says, you better do what I say, you're going to burn your house down, and then they burn your house down, why would you do what they say? And so, you know, I am from Canada, like all the best Americans. And in Canada, our response to the Trump tariffs has been incredibly stupid. We're going to tariff American goods, which is to say we're going to make them more expensive for us. Right. This is like punching yourself in the face really hard and hoping the downstairs neighbor says, ouch. Right. It's really dumb and it's indiscriminate. You're punishing, you know, and you don't take soybeans. You're just punishing some farmer in a state that begins and ends with a vowel who has never done anything bad to Canada. Meanwhile, you have these like eight tech bros who paid a million dollars each to sit behind Trump on the dais, and they extract billions from Canadians by having devices we're not allowed To modify, right? Well, we passed Bill C-11 in 2012 because the U.S. Trade representative said it was that or tariffs. It was wildly unpopular at the time. The two ministers who consulted on it got 6,192 consultation responses opposing it, 53 in favor. The main minister, a guy called James Moore, gave a talk to the International Chamber of Commerce where he said 6,192 responses were from, quote, babyish radical extremists, and he Was going to throw them away. No one likes this law. We get rid of that law rather than being like performatively angry about Elon Musk doing Nazi salutes, which just gives him the attention he craves. We can make it legal for every mechanic in Canada to jailbreak every Tesla in Canada and unlock all of the software subscriptions and upgrades without giving any money to Elon Musk. So we can create this gigantic consumer surplus in Canada, right? Make things in Canada cheaper. We can starve one of the guys who was the principal architect of Trump's regime of cash that he needs. And we can create an export product that everyone in the world wants to buy. You know, there was an era where Canada made the smartphone everyone wanted. Well, we don't do that anymore. Research in motion is a distant memory. But there are so many lines of business that have been off the table for America's trading partners because it was a condition of not having tariffs. And Canada isn't the only country in the world that has a pool of skilled technologists and capital. In fact, a lot of countries in this world are absorbing skilled technologists who have either been sent back by America or decided they don't want to live here anymore. And there are a lot of investors who would like an investment environment in which your success is not predicated on how many Trump coins you buy. And so there is a potential right now. And the first country that breaks, I think, is going to realize not just a consumer surplus, but they are going to consume the accumulated rents of American monopolists. ([Time 0:49:12](https://share.snipd.com/snip/0265c48c-2928-434e-9f9f-3ace4fc3bb56)) - **Limit Data Collection Now** - Stop collecting unnecessary personal data and enact privacy laws to prevent future breaches and discriminatory uses. - Cory urges immediate legal limits on data collection because past data breaches are largely irreversible. Transcript: Lina Khan What does the theory of inshittification say about how we stop these pieces of trash? Yeah. Cory Doctorow So, you know, as I mentioned, I'm Canadian. Down east in Canada, there is a joke whose punchline is, if you wanted to get there, I wouldn't start from here. And this is the situation we find ourselves in with privacy. And it's a thing that we've been saying as privacy advocates for a long time, that this data once collected is almost certainly going to breach, and it's probably never going to be destroyed. And once it breaches, it floats around forever. And there's so many ways it can be used. You know, one of the things that I think is wildly underappreciated is the risk of what's called re-identification. So a lot of data sets are, quote, de-identified, right? So they strip out some kind of identifying marker and they hope that that's going to keep the data from being identified with specific users. And there's good reasons to want to do this, right? If you want to do population scale interrogation of data about epidemiology, for example, you want to get a big de-identified data set and say, well, what are the correlates of certain COVID effects or who's getting the flu or whatever. And the problem is that when you have a breach, you can have a key that unlocks a data set. So say you release prescribing data for hospitals, right? But you use numeric identifiers in place of names. So you can see how the prescriptions go back and forth. You can see which doctors write which prescriptions. You can see, you know, is there an oxy epidemic? All kinds of things, right? But now say Uber breaches, and you know who went to the hospital right before those prescriptions were written. And you can now re-identify all the data and the data set. My friend Ben Goldacre, who's an evidence-based medicine specialist in the UK, often talks about this in the context of NHS data anonymization. Because we know when Tony Blair had a heart attack and we know when he had his heart surgery and we know where he had his heart surgery. We know how old he was when he had that heart surgery. We know how tall he was when he had it. We know how much he weighed when he had it. Any anonymized data set from the NHS that includes all the heart surgeries will immediately tell you which one of those people is Tony Blair. And you can find all the other data points associated with it. So we have been very cavalier about data collection. Many of you are probably Discord users. Discord was collecting driver's licenses as part of a public safety thing to know your user. They just breached them all. And we know that there was a breach of every social security number. We know that Equifax breached all of our data. Hard to overstate how toxic and how much debt we are in as a result of the failure to do privacy law. And the best time to have made a privacy law was when Die Hard was in theaters, but the second best time is now. I don't know what we do about all that data that's floating around out there. It's terrible. I mean, I can't overstate how bad it is that all this data is floating around out there. And I don't know what we do about that data in the past, but I know that we should stop it from flowing in the present and into the future. ([Time 0:54:44](https://share.snipd.com/snip/056b76a8-3e27-4080-bbb4-7feec035aaa8)) - **AI Art Lacks Genuine Intent** - Generative AI art lacks a communicative 'intender' and thus often feels hollow despite technical impressiveness. - Cory argues AI outputs dilute human intent and cannot substitute for meaningful artistic communication. Transcript: Cory Doctorow And Matthew Inman, who does the Oatmeal webcomic, just published something that I had written about this, but he published a webcomic the day before yesterday called something like An artist's reflections on AI art. It says it better than I can. So I think that art is a communicative act, right? Art is the process by which someone takes a big, inchoate, irreducible, numinous feeling that's in their head and they try to infuse it into some intermediary medium, whether that's Like a dance or a song or a painting or a book. And then they hope that when an audience member experiences that, that a facsimile of that numinous irreducible feeling materializes in their mind. And when you prompt an AI, you might have a big numinous irreducible feeling, but the only thing about it that the AI knows is the prompt. And so if you're writing like two sentences, then you're diluting two sentences across a million pixels, right? So like the actual like communicative intent in the resulting work is like it's homeopathically diluted, right? And, you know, there's a reason that we call this stuff like soulless or uncanny, right? Mark Fisher says, eeriness is the seeming of intent where there is no intender. We're not accustomed to finding paintings that don't have painters. We're not accustomed to finding words that don't have writers. And we do know about seeing faces in the clouds, right? But when we see a face in the clouds, sometimes we're struck by its aesthetics. We are not moved by the communicative intent of the clouds, right? We don't have that relationship to it that we have to art. We have a different kind of relationship to it. And I think that at first, AI art was very striking because we imputed intent to it. But with repetition, we've become inured to that. We've realized there is no intender. And it's become quite hollow, even if it's still sometimes striking. It literally has nothing to say. So I have a colleague who's a law professor, which you may relate to. And before Gen AI, before LLMs, people were very reluctant to ask her for a letter of reference because it was so much work. It's tedious. But now everyone expects you to write a letter of reference because they know that you put three bullet points into an LLM and it shits out five florid paragraphs. And so, of course, when she advertises for a postdoc, she's inundated with reference letters, which she then feeds into an LLM to turn into bullet points. And this is very funny and you can understand the bullet points are the same. But if you think about it for a second, the most interesting thing about this is that the AI knows nothing about the candidate. So everything that's not those three bullet points cannot be germane to the candidate. In the same way, everything in the picture or the words that the Gen AI generates is not germane because the AI has nothing to say, right? It's a word guessing program. It's a pixel guessing program. And so while a pile of leaves or formation of clouds in the sky can be striking, they are not communicative. They're not doing the work of art. And while the paragraphs around the three bullet points can seem like the kind of thing that indicate that a professor thinks highly of a law student, the reality is it is not germane To that law student. It tells you nothing. In fact, you probably can't locate the three bullet points that are germane to that law student. It is as though you threw a pile of leaves in the sky and when they fell to the ground, they formed a letter of reference. ([Time 1:01:43](https://share.snipd.com/snip/1231b1c1-ebba-43d2-9cc6-7d387e377f26)) - **EU Strength Hampered By Enforcement Holes** - Europe enforces tech rules more aggressively but federalism and tax-haven states like Ireland weaken enforcement in practice. - Cory highlights Max Schrems' litigation as forcing enforcement forward despite regulatory capture. Transcript: Cory Doctorow So the EU, it's an interesting situation. They are much more muscular about regulating U.S. Big tech. Some of that is that the companies are called Google and Facebook and not Nokia and Olivetti, right? And so there's just the political economy is that no matter how much Nick Clegg, like, who is Facebook's chief shell in Europe, insisted that Facebook was defending European cyberspace From Chinese communism. Like, nobody took that seriously. No one believes that Apple is an Irish company. And so they were willing to do stuff that was a lot more muscular, but they are hamstrung by one of their great strengths, which is federalism. And the great strength and the great weakness of federalism is that the states within a federation have a great deal of autonomy, which allows them to experiment and tinker in the margins And for there to be a variety of legal regimes. But it also means that one state can be your weakest link. And actually, Europe has got several states that are their weakest link. The big one is Ireland, but also Malta, Luxembourg, Cyprus, and so on. And the thing that these countries all share is their tax havens. And if you're a tax haven, you have to be a crime haven too. Because any company that can pay lawyers to pretend to be Irish this week can pay lawyers to pretend to be Luxembourgeois next week. And as a result, if you enforce any of the other laws, they're going to move from Dublin to Cyprus, right? And so the European privacy law, the general data protection regulation, it's now nine years old. The enforcement system for it begins in the national court of the country where that company is headquartered, where it's domiciled. And for surveillance, American surveillance companies to commercial surveillance companies, that's always Ireland. Now Europe had a bunch of crummy surveillance companies too. They didn't have the resources to pretend to be Irish. They were like, there's a mass extinction event. They all just went away when the GDPR kicked off. But the Irish data regulator is so captured that they don't get out of bed most days. And when they do, they sit around in their underwear eating breakfast cereal and watching cartoons. So we have these privacy cases that have gone to Ireland and died there for like nine years. And all those cookie consent dialogues that you get when you're accessing European Cider, if you're in Europe, we've got a VPN that has an endpoint in Europe. Those are illegal, right? They're just supposed to let you click past them and just keep going and not collect any data on you. And the reason that they're still there is because Ireland, right? Now, the GDPR had this glaring flaw. The new successor to it, the Digital Markets Act, moves primary enforcement for privacy cases to the federal courts. Now, whether that provokes a constitutional crisis remains to be seen, because we know in this country where our federation is a lot further along, that moving powers from the states To the feds is controversial for good and bad reasons, right? There are lots of states that don't like the idea of federal regulation, not because they want to, I don't know, keep black kids from attending school in Selma, but because they want To keep abortion legal and they are jealously guarding their powers. That might provoke a constitutional crisis in Europe, but so far so good. And this rule is coming into effect just as these cases that have been stuck in Ireland have finally exhausted all the appeals and are already moving to the federal court mostly due to One cranky Austrian guy named Max Schrems who basically is like a one-man tsunami of big tech litigation and he was like a law student at Stanford and he was learning about European tech Law while he was in America. And he was like, wait, I think this is all illegal. And, and so he started filing cases. It was how we ended up with the GDPR was, was Max Schrems, like trying unsuccessfully to enforce prior European privacy law. And now he runs a nonprofit called NOYB, which stands for none of your business. And, and he's doing all this impact litigation. It's finally moving to the European courts, the federal courts. ([Time 1:05:36](https://share.snipd.com/snip/7f969e0b-04cc-4796-b8db-b7e87b9919fb))