Transcript of our Conference Call on “The Antitrust Case Against AI Overviews” with Madhavi Singh

On October 22, The Capitol Forum held a conference call with Madhavi Singh, Deputy Director of the Thurman Arnold Project and Resident Fellow at the Information Society Project at Yale Law School, to discuss her paper The Antitrust Case Against AI Overviews. She joined Teddy Downey, Executive Editor of The Capitol Forum to explore how generative AI tools like Google’s AI Overviews may be reshaping competition in the digital ecosystem.

The full transcript, which has been modified slightly for accuracy, can be found below.

 TEDDY DOWNEY: Welcome to our Conference Call on “The Antitrust Case Against AI Overviews.” I’m Teddy Downey, Executive Editor here at The Capitol Forum. And today’s guest is Madhavi Singh, Deputy Director of the Thurman Arnold Project and resident fellow at the Information Society Project at Yale Law School. Her research focuses on antitrust regulation of digital markets, the economic and non-economic effects of monopoly power and consolidation in the AI supply chain.

Before joining Yale, she worked as an Antitrust Associate at Shardul Amarchand Mangaldas & Company in India, and as a Lecturer and Researcher in Singapore and India, holds a lot of degrees from the National Law School of India, University of Oxford, Harvard Law. Wow, very impressive stuff. And we’re excited to be talking to her about her recent paper, “The Antitrust Case Against AI Overviews”, exploring how generative AI tools, like Google’s AI overviews, may be reshaping competition in the digital ecosystem.

Madhavi, thank you so much for doing this today.

MADHAVI SINGH: Thank you for having me.

TEDDY DOWNEY: And if you have questions, as usual, throw them in the questions pane over here in the control panel. Or you can email us at editorial@thecapitolforum.com.

Madhavi, would love to just start off how do you think of this paper? How did you come up with the concept? And then maybe we could get into sort of the highlights from there.

MADHAVI SINGH: Yeah, glad to do that. So, this idea, in many ways, is not really new. Because I think people who’ve been working in the space of platform regulation have been privy to the conversation about how Google has been eating the lunch of publishers. Publishers have been complaining about this for a very long time, that they are heavily dependent on Google for referral traffic. And they have very little visibility or control over how their content is curated, ranked, presented, but also how much traffic they receive, and what percentage of the ad revenue from Google they receive.

So, in many ways, this isn’t really new. Even the use of publisher data or excerpts from publisher content by Google, that had been happening for a while now. Google had all these features like knowledge panels, which appear on the right-hand side of the Google search results page, and there are featured snippets and Q&A, where it had already started using publisher excerpts without their consent or without additional compensation. But when AI Overviews and AI mode was announced, it seemed like it was the final nail in the coffin, so to speak.

And even when it was introduced, a lot of publishers predicted that this would further decrease click-through rates and result in an increase in zero-click searches, result in a decrease in traffic, and an associated decrease in ad and subscription revenue. A few months after these features were introduced, a lot of reports and studies started coming in that publisher traffic had reduced by 30 percent, sometimes more. And there have been drastic decreases in ad revenue as well, which was obviously negatively affecting a lot of publishers.

So, that was the motivation behind the paper. I realize that this is not really something new, but it was a moment when we needed to act. There was a greater urgency to it because it had become really existential for publishers.

TEDDY DOWNEY: I know you say it’s not new. I mean, I think some of the harms are certainly not new. But you’ve applied this Section 2 framework, in a very detailed way, specifically to Google AI. And maybe you can walk us through, just on a high level, why this is a compelling case. And then we can kind of get into the mechanics from there. But we’ve got, I’m sure, people from DOJ and FTC listening in here. Why is this a priority? Or why is this a particularly good application of monopolization law?

MADHAVI SINGH: Yeah, you’re right. I think the harms were evident. But what was needed, or I felt was needed, was explaining how this can be captured by existing theories of harm.

So, to unpack that, I think it’s useful to think about what is the relationship between publishers and Google? The original bargain on which the internet was based was that publishers allowed Google to crawl their websites and to create a search index. And in return, Google sent traffic to publishers, which then publishers used to monetize their content.

The original bargain of the internet was referral traffic in return for publisher data. But as Google’s power over referral traffic started increasing, it started extracting more for less, which is classic abuse of monopoly power.

So, how did it extract more? What changed? I think how it started extracting more from publishers was that it started using publisher data in all these additional ways: for AI training, AI grounding, for summarizing publisher content in AI overviews, in Gemini and its other chatbots, and also for republishing publisher content directly in AI overviews and some of its other search features.

There was all this extra extraction that was happening because Google had this monopoly position in referral traffic and in search. While it was extracting more, it was paying less because now publishers were getting less traffic and less ad revenue. And to me, that seemed like a classic abuse of monopoly power.

To be exact, it’s actually a monopsony, not a monopoly. Because, while monopoly is a market in which there’s a dominant seller, monopsony is a market where there is a dominant buyer. In this case, Google is acting as a monopsony. It is the dominant buyer of publisher data, which allows it to buy this publisher data at lower than competitive prices.

And abuse of monopsony power, like abuse of monopoly power, is a violation of Section 2 of the Sherman Act. So, that briefly is the theory of harm that I think sort of captures what Google is doing to publishers here.

TEDDY DOWNEY: And I want to get into the exclusionary conduct in a second. But before we get there, I want to take a quick step back. Because something that occurred to me when I was reading your paper, and when you were just speaking there, when Google search started, the courts were much more favorable to Big Tech. Google got a lot of really good copyright rulings on scraping, on Google library. I forget that exact case, but when they were putting the books, Google books.

And to me, and you get into this a little bit. But the courts have seemingly, even though they’re mixed on AI and copyright violation, I think the way that Google is acquiring the information is much more legally dicey. Or is it fair to characterize it that way? Because obviously, they got all those favorable rulings on search when it came to setting up the crawler and scraping. But this is really a different mechanic in some respects.

And I’m wondering if that foundation, and where the courts are, make this a little bit actually easier of a time to make this type of case. Or is that not relevant? Because you do mention that some of the courts actually have been allowing fair use. But I would say it’s decidedly more mixed or problematic. We did a recent copyright conference call on AI that has me a little bit more pessimistic about Google’s chances of AI getting away with training and publishing when it comes to copyright. But before we get into some more mechanics, just want to get your reaction to that.

MADHAVI SINGH: So, I have mixed feelings about it too. I think that some of the copyright rulings, which were viewed as fair use, make me a little bit less optimistic. At the same time, I feel that this is distinctive. This is a distinct violation because it’s antitrust. It’s about having monopoly power and exploiting that monopoly power. And perhaps in some ways, there will be spillover of how those IP cases have been determined. But courts have had a few favorable rulings when it came to abuse of monopoly power or recent Section 2 monopolization cases.

So, it’s difficult really to say which way it’ll swing. But what makes me optimistic is the idea that there is an acknowledgement that Big Tech power is problematic and therefore we cannot allow Google to illegally monopolize all of these markets.

I think another good thing is that courts really struggled to find remedies in the Google search case. And it was also because it’s difficult to find remedies and restore competition in digital markets once it is lost. And I think experiencing that probably should make courts and regulators act faster because they know that competition once lost in these new markets is very difficult to restore. And hopefully, that will nudge them to act on these. But my guess is as good as yours on this.

TEDDY DOWNEY: Yeah, I think that point comes through in the piece also, that if we’ve learned anything, getting ahead of this monopoly exercise and monopoly power is better than trying to solve it after the fact. And maybe we could get into a little bit of the mechanics here of what a case would look like. Obviously, if the DOJ and FTC is interested, they’re going to do a whole investigation into this.

But you talk about exclusionary conduct here and it’s really how Google affects publishers. Can you walk us through that exclusionary conduct? We’ve mentioned it already, but what ends up happening to the publishers based on that exclusionary conduct?

MADHAVI SINGH: Yes. So, the exclusionary conduct is using publisher data without their consent and without compensation in all these additional ways and also republishing their content without consent or compensation. That’s the exclusionary conduct.

And how it affects publishers is that there have been reports of decrease in click through rates, increase in zero click searches on Google and an associated decline in both traffic and revenue. And I think this has consequences for the broader media ecosystem. We’re already in a media ecosystem where there are trends of consolidation.

Local news is struggling to survive. The fact that the ad revenue of publishers is drying up even further should be even more concerning. And I think, just to sort of drill in on the harms even more, I think it’s obvious how publishers are being harmed and how that might affect media, information generation, and even democracy, for example.

But I think users will also be impacted. And I think this is a point that is much less discussed because right now we have the option to choose between so many different publishers. There’s New York Times, the Wall Street Journal and also some local publishers.

But as the costs of operating for these publishers increases and ad revenue dries up, a lot of these publishers will exit the market. So, users will have less choice of content. And publishers will be able to invest less in journalists and in editors, which would also affect the quality of content that users get.

And one could even go further. This will have effects on even AI chatbots. So, right now AI chatbots have so much training data because publishers are producing original content which they can train on. But as publishers exit the market, the original content available for training AI models will decrease, which will also affect the quality of AI models. The quality of AI models will either plateau or even deteriorate as publisher content disappears. So, it could have spillovers on so many different markets. But I think that there is a very clear case to be made of harm on both publishers and consumers.

TEDDY DOWNEY: Also, not just the models themselves, but often the models, when you put in the chatbot, the model will then go do its own Google search to get more information. And they’re going to be searching your sites that you have fewer sites and you’re going to get a worse outcome from that. So, you also mentioned not only that Google is scraping websites without consent or compensation to train the models. You mentioned publishers have no recourse to resist this.

MADHAVI SINGH: Yeah.

TEDDY DOWNEY: And Google’s critical for these companies to get discovered and to distribute. Can you talk a little bit more about that? I think this is kind of glossed over sometimes that these are critical distribution channels and there’s no real recourse for the publishers. Maybe we can talk about that for a second.

MADHAVI SINGH: I think it’s very interesting because this came out during the Google trial. There was this internal document that was released during the Google trial where Google was discussing whether and how it should use publisher data for AI training and what were the different options that it could give to publishers. So, should it give publishers the option to opt out of their data being used for AI training while still allowing them to be indexed on search?

And they decided against it. There was a period when Google gave publishers the option that their data would be used only for search index, but not for AI training. But they never ended up honoring that promise. And all of this came out during the Google search trial.

So, currently as it stands, publishers don’t really have any recourse to opt out of their data being used for AI training. If, as a publisher, you want to be on the search index – and, of course, you want to be on the search index to be discovered ‑‑ then you would also have to agree to your data being used by Google for all these AI training purposes. So, that’s a sort of a tie-in. And publishers don’t really have an opt-out option. And even when there was an opt-out option, Google did not honor it.

There have been some companies that are trying to find workarounds, but it’s really difficult. For example, Cloudflare, early this year announced that it would block AI crawlers by default, at least for the publishers that it handled. And publishers could then choose if they want to allow AI crawlers.

Once again there we are relying on this private company, which has a lot of power to do this for us, when instead we should have better ways to take care of this. But currently publishers at least have no way to opt out of their content being used by Google for training its AI models, if they want to be on the search index. So, it’s either all in or all out.

TEDDY DOWNEY: I want to get more into tying in a second. There is another aspect of this exclusionary conduct that I want to talk about, which is not attributing properly, or misattributing as you describe it. And maybe we can talk about that. Is there a lot of evidence that happens? And how is that problematic from a legal standpoint?

MADHAVI SINGH: Yeah, so there is early evidence of how Google has started attributing less content. So, it started out with more links. Now it has less links on its AI Chatbot, Answers.

Additionally, it sometimes uses content without actually attributing sources at all. And once again, publishers have been at the forefront of showing that this happens anecdotally. There are also some studies which show that this happened. But it’s difficult to see how many times Google has actually done this.

But we do see evidence of Google using content without attributing it, especially in the context that you mentioned, when, apart from just like usual AI training, when chatbots go out and look for content through search. When you look for something more contemporary, like a news item, and chatbots go out on the internet and start looking for that content.

In terms of how that’s a problem legally, I think it’s once again abuse of monopoly power ‑‑ or abuse of buyer monopoly power – it does not just take the form of extracting monopoly rents or higher prices or lower than competitive prices. In this case, it’s not just price-oriented, but also quality-oriented.

It’s about if you, as a monopoly, can impose more exploitative terms, that is also an abuse of monopoly power. And in this case, I think, imposing these terms, not attributing when you’re taking content from publishers, that is an exploitative term, which is an abuse of monopoly power.

TEDDY DOWNEY: And then you also mentioned, Google prominently placing AI generated answers at the top of search results, making that more and more and more of what the user sees. How is that problematic?

MADHAVI SINGH: Yeah, so over time, Google has been increasing the percentage of the screen on which AI overviews appears. And this is especially problematic when you’re viewing things on a mobile phone, and you have a smaller screen. Because most of that space is taken up by AI overviews. And, again, we know that if something is at the top of the search results page, it’s more likely to get clicks.

And it’s the same with AI overviews. The fact that it is at the top and it’s also prominently displayed in this sort of separate box, it means that most users now read what’s in the AI overviews. And if it’s just a small query, a lot of users don’t click on even links that are present in the AI overviews, forget scrolling down to the remaining search results page and clicking on some of the organic links down there. So, yes, definitely having AI overviews at the top, and also prominently displayed further exacerbates the problem.

TEDDY DOWNEY: And we’ve already mentioned this, but now that we’ve kind of gone through all the exclusionary conduct, and we’ve talked about some of the impact on publishers — obviously, traffic is massively down. You hear about publisher layoffs all the time. You already mentioned the decline of newspapers and threat to democracy and all that. Are there any other anti-competitive effects that we should think about that this case would bring up, this hypothetical case could look at?

MADHAVI SINGH: Yeah, so I would like to point to two additional things. I think that this type of conduct also helps Google maintain its monopoly in the general search services. And we know that Google already has a very strong monopoly, it has had market shares of between 88 to 90 percent over the last several years now.

This type of conduct allows Google to maintain its monopoly position in general search services. How does it do that? Because Google is a monopoly buyer of publisher data, like we discussed, it’s able to get this input for lower than competitive prices.

But none of the other search engines like Bing or Yahoo, or DuckDuckGo, or even new entrants, because they don’t have the same market power as Google, they would not be able to get this publisher data at these lower-than-competitive prices that Google is getting it for. And so, it makes it difficult for its rivals or even new entrants to compete with Google on a level playing field in the market for general search services. This conduct helps Google’s monopoly maintenance in general search services.

Apart from that, I think it’s also useful to think about this problem just more broadly, and to think about Google’s position in a bigger ecosystem. Google, of course, is at the center of this ecosystem. It controls so many different apps like Maps, YouTube. And it also has relationships with third parties like Apple and Samsung. And Google can leverage all of these pre‑existing relationships with third parties, in ways that ensures that it’s attempted monopolization of AI chatbots, et cetera, actually becomes successful. And we see this happening.

In the aftermath of the Google remedies decision, once Google knew that Chrome was not being divested, it immediately put Gemini on the Chrome browser. And it is now bundling Gemini with all of its apps. You will see Gemini very soon on Maps and YouTube and all of Google’s other apps.

This type of conduct where you allow Google to use publisher data in all of these ways will have spillover effects. Because Gemini will not just stay as AI overviews on Google search, but it will also be used across a suite of different apps, which would make it even more difficult to unbundle when you are trying to think of a remedy. And it will have spillover effects in so many more markets. It would entrench Google’s position.

Similarly, with third parties, Google has started doing something very similar to what it did with search. Google has started entering into deals with Apple and Samsung to put Gemini onto their smartphone devices. Another classic thing from Google’s playbook.

Google’s position in this ecosystem is just so strong that it will have effects across a range of different markets. And a range of different consumers will be affected.

TEDDY DOWNEY: Before we get to the pro-competitive justifications, the defense from Google that could crop up here, one of the things we saw was that the people who were dominated by Google in the past, publishers Yelp, they were very loud complainers. In this instance, when Google’s just tying it across all their entire ecosystem, who is going to be harmed in addition to the publishers? Like, what’s the Yelp of this strategy? Is it Open AI? Is it the other chatbots? Who are the people who are going to start complaining about this conduct?

MADHAVI SINGH: So, I think it’s unlikely that other AI companies would bring antitrust suits. Because as monopolies of the future, you might not bring cases against monopolies of the present. History will not remember you very fondly for that. But I think that there are some Yelp equivalents.

For example, in the U.S., there are already two complaints that have been brought on similar theories of harm. And one is by Chegg, this online education company, which claims that Google ends up using most of its content to answer students’ questions in AI overviews, and then no one clicks on the Chegg website. And because of that, it has had a drastic decline in revenue. So, you certainly have these companies which are Yelp equivalents. And, of course, you have traditional publishers who don’t have a lot of bargaining power vis-a-vis Google, but they’re still big, well-resourced players. For example, Penske Media, which is the owner of Rolling Stone and Hollywood Reporter and a few other publications, also brought a complaint against Google on a very similar theory of harm against AI overviews and AI mode.

So, we have these private litigants. But I think that because of the scale of the case, and because I believe it implicates the future of the internet, it is insufficient that you just have private litigants bringing cases. I think given the scale of this, the federal regulators, or at least the state AGs should get involved.

But I’ll stop there. And we also mentioned the success and what’s the likelihood of success, et cetera. But I’ll speak more about that in case there are more questions. Or maybe we can touch upon that later.

TEDDY DOWNEY: Yeah, at the end, I want to talk about why DOJ or FTC should at a very least launch an investigation and take these arguments very seriously. But before we get to that, you make so many good points in the paper, I just want to kind of get to a few more of these.

Google is obviously going to come up with their defense, their justifications for this. Again, I think a lot of this goes back to just the framing of the Big Tech of like, we should be allowed to do whatever we want. And that’s kind of carried the day for them. Government shouldn’t be able to tell business what to do, that type of thing. But if you could lay out what Google’s justification for this is. And then maybe we could get into tying after that.

MADHAVI SINGH: Yeah. So, at least right now, there is little publicly available justification, but Google claims that AI overviews actually helps publishers. It allows a more diverse set of publishers to have access to the eyeballs in the beginning, because it’s possible that some not so well-known publishers might be featured in AI overviews, et cetera.

So, it just denies the idea that having AI overviews in search results actually results in decline in publisher traffic, whereas we have a lot of studies showing to the contrary. So, that’s Google’s claim that it just denies that publishers are negatively affected because of this.

But I think when this becomes an antitrust lawsuit, we will also have Google making more technical claims, like trying to argue that this is actually just a product design. It’s a design improvement rather than two distinct products that are being tied together. So, there’s no tying. If you just improve your product in some ways, then that’s not a distinct product that you’re tying together with something else. But that’s not an argument that Google has made just yet. I think that once an antitrust suit is brought, Google will probably make those arguments then.

TEDDY DOWNEY: Yeah. We’ve seen Google argue, depending on the venue, oh, it’s the same thing. Oh, it’s different from search or AI, depending on where they are. So, it seems like you agree that is sort of an important thing to keep an eye on how they position themselves around is it distinct product? Is it tying? That is worth keeping an eye on. Now, when it gets to tying, that’s obviously another aspect of what a case would look like. How do you see the facts of what Google’s doing here fitting into tying allegations?

MADHAVI SINGH: I think that there are two ways to see this as a tying violation. One is Google putting AI overviews at the top of the search results page or Google tying AI overviews and Gemini with Google search and also bundling it with its other apps. That’s one type of tying where I think that there are clearly distinct products. Because Google had its search engine and there’s this chatbot, which are distinct products.

We know Google and Bing and others operated search engine as a distinct product until recently. And then there are obviously the chatbot companies as well that operate just the chatbot. So, they are clearly two distinct products, which are being tied together rather than just one product of Google Search, which has been improved through new product design features. That’s one type of tying that is happening, tying or bundling of AI overviews and Gemini with the search results page and also with Google’s other apps.

But another type of tying, and I think this at least was very clear to me, was just the tying of referral traffic with allowing publisher content to be used for AI training as well. And we talked about Google’s monopoly power as a buyer. As a buyer, it is tying access to referral traffic for publishers with using publisher data for AI training purposes. That’s another type of tying. And I think that the evidence of that is very clear.

TEDDY DOWNEY: And from a legal standpoint, you mentioned ‑‑ and maybe this is kind of good for our international audience as well ‑‑ you mentioned that the standard for proving tying is a little different in the U.S. versus the EU. Maybe when it comes to requiring proof of coercion, maybe you could walk us through that. That’s kind of an interesting distinction in U.S. and EU law, maybe allows the EU to potentially be a little bit ahead of the U.S. potentially, at least on this front.

MADHAVI SINGH: Yeah, that’s very interesting. I think because some parts of this case, like tying of AI overviews to the search results page, that to your European audience might sound very similar to the Google comparison shopping case, which was one of the first competition cases, to be brought against Google and Alphabet in the EU. The complaint was basically that Google had tied its own comparison shopping website to the top of the search results page, and that disadvantaged all other comparison shopping websites. So, they’ve already been thinking about this, and their theories of harm like tying, et cetera, have space to accommodate this. Whereas, we don’t really have that in the U.S. so far. Because there have been very few tech monopolization cases in the U.S.

Google obviously is the biggest case that we’ve had in recent years. And Google also wasn’t based on tying. It was illegal monopolization, Section 2. It wasn’t based on tying. So, we haven’t really tested this in court about what tying would look like in digital markets. And whether the conduct of having AI overviews at the top of search results page and tying that, whether that would fit within tying doctrine in the U.S. I think that’s still an open question. And there’s not really much that has been decided on it. But I think that the EU is a good place to look at. Because they’ve already figured out how this can be brought within the framework of tying.

TEDDY DOWNEY: You talk about some of the challenges in the U.S. But are there any big high-level reasons for why you think the antitrust enforcers have steered clear of making tying allegations when it comes to these tech platforms?

MADHAVI SINGH: So, I think that the Google search case was very limited. I’m not sure why they didn’t bring tying allegations, et cetera. There were a lot of allegations that could have been brought against Google, including for its app stores, and for how Android was being operated, and Google suite of apps, et cetera.

But they decided just to bring a case based on search and exclusive distribution agreements that Google had entered into with these third parties. I think that could have been a strategy, because it’s easiest to prove. And I think it’s easiest to show that kind of exclusive distribution agreement is illegal. But I’m not sure. And there is much to be answered on what tying would look like in the U.S. in digital markets.

TEDDY DOWNEY: And then, I think a lot of people think about Microsoft as the big case when it comes to monopolization. Obviously, there are a lot of ones that are going through the process now, but not a lot of established precedent yet. We’re still waiting on opinions and appeals and things like that. Walk us through how you see this being similar or different to the Microsoft case, because that is still how a lot of people think about monopolization cases today.

MADHAVI SINGH: Yeah, so in the paper, I referred to this as the new Microsoft moment. And I was a bit worried even writing that. Because I felt like it’s such a cliche to invoke Microsoft at the drop of a hat every time we talk about antitrust regulation of platforms. But like you said, it’s because Microsoft is the biggest antitrust monopolization case that we have. But I do think that this moment is similar to the Microsoft moment in principle. And I think that’s because Microsoft, at the core of it, was about a company that had a monopoly in one platform, and preventing that company from becoming a monopoly or gatekeeper for the next generation of technology.

So, Microsoft already had a monopoly in operating system, and operating system was the site or platform where most user activity took place. Mostly people used apps for listening to music or for productivity tasks, et cetera. But the threat of the internet resulted in Gates writing the famous Gates memo about how people are going to move from using operating system to the next layer, and most of these tasks would now be done, not on domestic apps but on the internet through browsers. And so, they were afraid that they would lose this gatekeeper position of being the platform where most user activity takes place.

I think that this is a similar moment because Google declaring Code Red to me is analogous to Microsoft’s Gates memo. Google declared Code Red when ChatGPT came out and became popular. It is analogous to the Gates memo moment, because I think for Google, it already has a similar gatekeeper position in general search. It is the gateway through which you enter the internet. But now it risks losing that gateway status to chatbots and AI answer engines. It is now trying to find a way to leapfrog from being just a search monopoly to also monopolizing the next gateway to the internet or to the next generation of technology. In principle, this looks very similar to the existential threat that Microsoft was facing back then.

TEDDY DOWNEY: It’s really interesting. And then I want to get back to when you were talking about why this is a good case for the DOJ and FTC to put resources into. We’ve already discussed the timing, right? This is great timing for a case because the monopoly hasn’t entrenched itself yet. It’s at the beginning stages of this new product, which is a more timely way to do antitrust. What are some other reasons why the agencies should look at this to your mind?

MADHAVI SINGH: I think that it’s because big innovation incentives and the future of the internet are at stake. Just like we had monopolies controlling the gateways to the internet in the past, we do not want to repeat those mistakes and not just have old monopolies controlling the new gateways. But also, we don’t want new monopolies controlling the new gateways to the internet.

A case like this would not just prevent Google from monopolizing the new gateways to the internet, but it will also provide a signal to the others, to the other attempting monopolies out there who want to control the new gateways to the internet, that this type of conduct will not be tolerated and that the DOJ and the FTC now want to act early and are looking at this very closely.

So, it’s both about preventing actual harmful conduct, but also about providing a signal to the others who will copy the playbook of Google, both in Web 2.0 and also what Google is doing now to do similar things and tell them that it’s not something that federal regulators or state regulators will tolerate. So, I think that it’ll have so many effects. It’ll affect innovation, and I already touched upon how it’ll have spillover effects in a lot of different markets. It’s not just about search engines. It’s also about AI chatbots. It’s about publishers and the future of publishers and ad revenue for them.

TEDDY DOWNEY: And you mentioned you wanted to touch on the likely success of this type of case. What do you think really makes it compelling? Obviously, you wrote a paper all about how the litigation would work. But just in terms of the evidence and the moment, where the courts are, why do you see this as particularly likely to be successful? And then we’ve got four questions here from our listeners to get to after that.

MADHAVI SINGH: About the moment that we’re in, I think that in some ways this is a very good moment to bring this case. Because I know that there’s an overall sense of disillusionment with the remedies that antitrust is capable of imposing, or even just the general regulatory and judicial appetite for imposing good, strong remedies in antitrust cases. But I think that this is a good moment to bring this type of case because even a very simple remedy here could go a long way.

So, by a simple remedy, I mean, even like a simple injunction which tells Google not to use publisher data in all these additional ways and to give publishers the option to opt out of their data being used in all these additional ways while allowing their data to be used for the creation of search index. I think that’s just a simple injunction. Even a simple remedy like that could go a long way.

I think another reason why regulators should be interested in taking up this case is because the remedy here is clear and it’s clear just because we’re acting early. Once the market does get monopolized, remedies to restore competition in digital markets are much more difficult to design. Or convincing a court is much more difficult, especially as courts have proven themselves to be so cautious and conservative.

Another reason for regulators to bring this case is because the theory of harm is clear. The remedy that you are going to ask for is clear. And because the remedy is so simple, it’s more likely that a court is going to grant it. So, I think for all those reasons, I think it’s a good moment to bring this case.

But at the same time, I’m not very optimistic that a case is going to be brought by federal regulators right now because of all the national security rhetoric that AI tends to get couched into. There is this phantom of the U.S.‑China AI arms race always looming in the background.

But for that sort of argument, I would say that one should closely examine the assertion that less competition or protectionism is good for creating strong American tech companies. I think more competition is better for creating strong American tech companies. And therefore, even from a national security perspective, I think subjecting your Big Tech companies to competitive discipline would help you win the U.S.‑China AI arms race, assuming that’s a normative goal you want to pursue.

TEDDY DOWNEY: We have two other questions. What do you think about the states and the EU? Obviously, the EU is considering their own tech stack interests now. And the states obviously, to some extent, have been showing signs of ramping up their interest in bringing antitrust cases, showed a lot of leadership on Google ad tech, for example.

What do you think about just the opportunity for the states or the EU to step in here? They may seem like they have different and opposing incentives from what you’re saying from a national security. I would call that the red herring argument or whatever you want to call it. Seems pretty patently disingenuous on many levels. But what do you think about when it comes to states and in Europe?

MADHAVI SINGH: In many ways this paper was written for state AGs. That was the audience I wanted to get to because I wasn’t optimistic that federal regulators would act on it because of the red herring national security argument.

State AGs have the same powers. And you’ve had excellent interviews in the past on The Capitol Forum talking about the role of states in antitrust enforcement. So, there is some discussion now about states taking on a more proactive role in antitrust enforcement. And I think states have a lot of power under state regulation and otherwise. So, they can use all of these same theories of harm. And they have their own state antitrust statute. And I think some instances, like in California, have shown us that state regulations have spillover effects. And sometimes when tech companies comply with state-level legislation and remedies, they would end up implementing that, not just in a specific state, but would do it for the entire country. So, I think it would have the same effect. And I’m hopeful that state AGs would act on it.

The EU, of course, has been very proactive and it has its own set of geopolitical tensions right now that it has to cater to. That being said, complaints have been brought by independent publishers before the European Commission on very similar theories of harm. So, they will be called upon to decide it as well. And apart from antitrust, because EU has been so active on the ex-ante front as well, I think EU could also explore more regulatory options.

For example, Australia has the News Media Bargaining Code, which targets some very similar things, ensuring that publishers can negotiate fair compensation with distributors and platforms like Google and Facebook. I think Europe could also explore similar options. I’m not very sure if they are already doing that, but given EU’s regulatory appetite for doing ex-ante regulation, I think it would be an interesting template to look at.

TEDDY DOWNEY: And we’ve got listener questions here. Here’s one. We know that Gemini and Search existed separately for a time. How do you think consumers think of them now? This seems particularly important for the claim they are two distinct products. We talked about this a little bit. But this is specifically asking about the consumer perspective when it comes to Gemini and Search. What are your thoughts on that?

MADHAVI SINGH: Yeah, that’s super interesting because I don’t know if distinctiveness of products and that assessment depends so much on consumer perception, but I’m sure it plays into the mix. And it also means that right now we still have consumers ‑‑ most consumers who use the Google Search now have seen Google Search without AI Overviews.

But your listener is right that as time passes, as these products get more and more integrated, it will be difficult to see them as being distinct, especially as a very large part of Google Search results page ends up being the AI Overview. So, that argument might be more difficult to make as time passes and people start perceiving Google Search as not just Search but Search plus AI Overview.

TEDDY DOWNEY: I think it also depends on what this user means by consumer. Obviously, the publishers are buying or otherwise selling. I guess they’re interacting with these products. So, they’re users in some respect. But he’s ostensibly thinking about an end user, I would imagine.

MADHAVI SINGH: And just a point on success also, I think that the whole monopsony story of Google being the monopoly buyer of publisher data and holding power there, I think that’s the stronger argument that one should run with, especially as we don’t know what the tying doctrine is going to look like in digital markets.

TEDDY DOWNEY: We’ve got another question here. How did this topic make its way, or could it have made its way, into the Google Search case remedies?

MADHAVI SINGH: So, I don’t have too many thoughts on it apart from one potential solution to this, which some people might find a little bit more extreme. But one potential solution to this is having a line of business separation or structural separation. Google already has a monopoly in search engine and so, maybe, it should not be allowed to enter this separate market of AI powered answer engines or AI chatbots, which is a structural separation remedy or line of business separation remedy.

That type of remedy could be demanded in a case like this. I think the only Google remedies decision and the Google Search case, connection that I can make here is that DOJ asked for a very similar looking remedy in the Google remedies trial phase, asking that Google not be allowed to invest in AI and that was rejected. So, maybe that makes such a remedy in this case also less likely to be accepted.

Apart from that, one way in which the Google case is relevant is that Google’s monopoly has been established. We’ve gone through the process of defining the relevant market and showing Google’s monopoly power and we don’t have to do that all over again in a case like this.

TEDDY DOWNEY: I have a question here that I’ll admit I do not understand. It just says illegal reciprocal tying. Is that just that we were talking about tying? Or is this reciprocal tying a term of art I’m not familiar with?

MADHAVI SINGH: I think that’s reciprocal dealing, but there’s also tying and I think this is tying. So, I’m not sure.

TEDDY DOWNEY: All right. We’re out of questions. Maybe the last question I have, again, is just to take a step back and think about this moment again. I think there are broader themes here with a lot of people questioning rule of law, questioning fairness in the courts, questioning power that these companies have over our lives. We talked about this, the threat to democracy. You briefly touched on this. But what are some of the bigger picture reasons to bring a case like this now?

I mean, you mentioned in the Microsoft example, that case really did deter Microsoft from extending its monopoly. Obviously, it’s still somehow endured as a monopoly to this day but at least temporarily opened the door for more competition. That case ‑‑ mergers were scuttled that Microsoft otherwise would have pursued. It was a big check on corporate power.

What do you think about this moment, these broader themes and the opportunity that this type of case would bring to at least address some of those concerns and fears about threats to democracy and rule of law?

MADHAVI SINGH: Yeah, so much to unpack here. But there is, I think, very quickly, two things: future of the internet and also democracy plus rule of law. So, on the future of the internet, I think it’s always sort of difficult to imagine what the counterfactual world would look like had there not been a Microsoft antitrust case. But a lot of people have very persuasively argued, that had it not been for the Microsoft antitrust case, all these amazing tech companies would not have emerged, which also ended up becoming monopolies in their respective markets. But you would have no Google and Facebook and so on and so forth.

But back in that moment, it was very difficult to predict what that company could have been, at least when the Microsoft trials were going on. Similarly, it’s very difficult to say now what technology would emerge if we prevent Google from monopolizing these markets. But I think the fact that there are so many AI startups that are putting in so much resources and money into finding alternative ways to access the internet means that there are some good ideas out there and we should give them an opportunity to win.

And if we allow Google to leverage its monopoly position in search to also monopolize these future markets, we would not be giving them any opportunity to succeed or a fighting chance. I think that it’s difficult to say what the future of the internet would look like, but the past has told us that more competition is better and so we should pin our hopes on that.

On the future of democracy and rule of law, I think that this is so important because a lot of what is at stake is not just digital markets, but publishers and journalism. And we already see that having fewer players in media also means that it’s easier for the government to control those media companies.

And therefore, I think it’s very important that we ensure that there are more and more publishers, there’s plurality of publishers who cater to different consumer bases. And that’s very important to ensure that the fourth pillar of democracy, which is the press and media, actually has a chance to survive.

TEDDY DOWNEY: That was a great answer to an impossible question. So, I thank you for that. We’ve got one last question from the audience.

Do you see there being a middle ground between publishers’ content either being used for AI overviews or not being used? And specifically, are there any workable monetization models that would allow for the continued use of AI overview while also protecting publisher revenues? The idea that we just stop Google from using AI Overviews seems like trying to stop an emergent technology. A bit of a leading question. Curious if this person works for Google. But go ahead, Madhavi.

MADHAVI SINGH: Yeah, I’ll try my best. So, even if AI overviews are good and they are consumer welfare enhancing, what publishers still deserve, and they are entitled to, is fair compensation. We still need to find a way to compensate publishers for their data that has been used in training those models and in creating their summaries or content in AI overviews.

Maybe what you do when you prohibit Google, when you impose an injunction in a case like this and prohibit Google from using publisher data in all of these ways, is that you force Google to enter into licensing agreements with publishers or you force them to figure out some other model of payment, for example, pay per crawl. So, Google doesn’t end up cornering all of those monopoly profits for themselves and some of it goes back to publishers who provide this content.

TEDDY DOWNEY: Yeah, as a publisher, the point I would make and I think you were right to suggest that there are companies that are going to be afraid of making a monopoly argument even if it’s in their best interest to have competition here. So, they could potentially win that competition on the merits as opposed to just defer to Google. But it is short-sighted in my opinion for these companies to decimate the publishing industry because they need publishers to train the data. It’s going to be a crummy product.

Like Google got crummier, got worse and worse and worse, because they just put ads on it. I literally can’t even use it because I can’t tell what is an ad and what is a link anymore. Maybe if they hadn’t done that, you wouldn’t be so easily replaced by a chatbot that is serially stealing copyright, stealing people’s content.

And obviously, in this respect, I am a publisher who vigorously defends our copyright. But I do think at some level it’s short-sighted that, look. If this product is going to be good and accurate and useful to the citizenry, it’s going to absolutely depend on a robust publisher industry. And in fact, you’re seeing the licensing happen already because they’re losing in court on the copyright claims. And to me, that’s really where the rubber meets the road. You can no longer have a license to steal content. You have to license it. And to me, this is another avenue of ensuring that payment occurs in a way that makes market sense.

And to me, there’s no such thing as a chatbot model with no publishing industry because they’re just so heavily reliant on those inputs being of a high-quality nature. You’re just going to have garbage, amateur and erroneous content, that filling the bot. And then that’s going to be useless. So, and then, you know what? That might open the door to a good search engine potentially. No.

But Madhavi, this is an amazing conversation. So, much to go through. So, many tough questions. So, many interesting things. I’m super excited to see how our audience of state AGs and policymakers at DOJ and FTC respond to this paper and how the courts look at it in the future, hugely influential work. And thank you so much for joining us today.

MADHAVI SINGH: Thank you so much for having me and for your excellent questions. Thank you.

TEDDY DOWNEY: This concludes the call. Thanks for joining us today. If you found the conversation helpful or thought-provoking, please be sure to follow The Capitol Forum on LinkedIn X and Blue Sky for breaking news, policy updates, and information on upcoming events. We’d love to hear from you. Drop us a line at events at tcfpress.com if you have ideas for future topics, speakers, or just want to connect with us. Have a great day, everyone. Thanks again for doing this.

Thanks so much, Madhavi.

MADHAVI SINGH: Bye. Thank you.

TEDDY DOWNEY: Bye.

(END OF TRANSCRIPT)