Katherine Forrest: Well, I have a third favorite thing, which is my dog, Milo. Actually, I have more than that. I also have my spouse, and then I have a few people I like. Maybe only a few. But anyway, it is two of my really super favorite topics. And so we’re going to be able to put my love of AI here together with the antitrust. And for those in the audience who don’t know that I was an antitrust lawyer, among other things, for a number of years, I was. I practiced antitrust and still do for years and years, particularly in the media and entertainment industries. And then also I did a lot of merger clearance in the day. I did the Continental United Airlines merger for United and tried that case and also got it cleared through the agency. So I did a whole bunch of those kinds of things, and I was an MDL judge for antitrust. So let’s just jump in and talk about the Cartwright Act. And Matthew, I think that is your job because you need to tell us why here on an AI podcast people should care about it at all.
Matthew Robinson: I want to tell you that I consider myself an honorary AI attorney now before I get into the antitrust of it all.
Katherine Forrest: Right.
Matthew Robinson: But I think it’s important that we first sort of set the federal baseline, which—so to start, we have to really start with the Sherman Act, which is sort of this long-held, it’s sort of one of the cornerstones of antitrust law, both Section 1 and Section 2. So Section 1 of the Sherman Act really focuses on conduct that is based on contracts, combinations or conspiracies that ultimately restrain trade. While Section 2, which is less relevant for this particular discussion, really discusses monopolies and attempted monopolization. So I think from a federal perspective, the Sherman Act will be most relevant to our conversation here, which leads us into the Cartwright Act and other state antitrust acts.
Katherine Forrest: Yeah, so tell us why we’ve got both a federal set of laws and state laws. And then tell us about the Cartwright Act.
Matthew Robinson: So if the Sherman Act is the cornerstone, these state antitrust laws are almost like its children, right? They kind of mirror and track the Sherman Act in many ways. So in many ways, you will see these pled together. So in different courts, they’ll have—you know, we’re on a case currently where they have the Federal Act leading the charge and then a plethora of state antitrust laws that have their own specifications but largely do track the issues covered by the Sherman Act. So for instance, in New York you have the Donnelly Act. And in California you have the Cartwright Act.
Katherine Forrest: So give us a quick overview of the Cartwright Act.
Matthew Robinson: So again, the Cartwright Act tracks the Sherman Act in the sense that it targets conspiracies that seek to restrain trade or foreclose competition. But it’s often described as its sort of broader or deeper or more wide-reaching counterpart to the Sherman Act. For instance, it includes indirect purchasers. Indirect purchasers can sue for damages, for instance, under the Cartwright Act.
Katherine Forrest: You’ve got to tell us what an indirect purchaser is. I think I know—I do know I know—but I’m not sure our AI listeners know. What’s an indirect purchaser?
Matthew Robinson: Well, when I first heard it, I did not know, so maybe it’s not so clear. But an indirect purchaser is—I think I can best illustrate it through an example. So if a company has a product, it seeks to sell its product first to direct purchasers, which are, for instance, the stores. The indirect purchasers are the consumers that then buy that same product from the store itself. So the stores would be the direct purchasers and the consumers that go to the intermediary of the store are the indirect purchasers.
Katherine Forrest: And let me sort of give you another example. Let’s just say that you are somebody who makes iron products and you buy iron ore as an input. So if you’re the first manufacturer, say you’re going to make a sheet of iron from the iron ore, you would be the direct purchaser of the iron ore. But if that same sheet of iron then gets sold down the line to, say, a boat manufacturer, then that boat manufacturer is an indirect purchaser of the iron ore. And so, you know, it goes on like that. So you can really take any purchase and sort of trace it and figure out how many layers it goes through. So I got it.
Matthew Robinson: And I just want to add: like the Sherman Act, it also allows for private trebling of damages.
Katherine Forrest: Okay, well that’s actually a big deal because trebling of damages means you take the damages that are assessed, maybe at the liability stage, and it gets multiplied times three.
Matthew Robinson: Yes, and that’s actually a very important part, right? Because we have clients all the time that encounter these sort of antitrust suits. And one of the biggest issues they have is when they do the damages calculation, it’s really three times, right? You can three times the award, which is really impactful, as one can imagine, to different defendants. But let me talk about the key concept that connects to what Scott will talk about, I think, in a moment here, which is allegations of price fixing. Because one of the things that an agreement or an understanding can relate to under the Cartwright Act or the Sherman Act is an agreement on price. Those are what we call per se violations of the antitrust laws and are also taken very seriously.
Katherine Forrest: Right, and price fixing, it’s happened from time to time. There have been and are lots and lots of claims of price fixing. There are, of course, claims that have no basis. In fact, they’re alleged nonetheless. And then there are times and instances in this world where people have price fixed. And that really means competitors getting together to fix the price of a product. So the classic example would be people getting together who sell a particular kind of product and they sit in a smoke-filled room and they get together and say, “hey, I’m going to sell this at five bucks, Billy. You’re going to sell it at five bucks?” “Yeah, Johnny, I’m going to sell it at five bucks too.” “Okay, Billy, if you don’t sell it at five bucks, I’m going to break your kneecaps.” “All right, Billy, I’m going to sell it at five bucks.” You know, how does that—you like my price fixing voice? All right, so Scott, give us a sense about where this AI piece fits into all of this.
Scott Caravello: Yeah, of course. So AB 325, which is the bill that put this amendment into effect, was signed on October 6th. And in amending the Cartwright Act, the author of the bill, who was the California Assembly’s majority leader, stated that it’s to modernize California’s antitrust laws to address algorithmic price fixing specifically.
Katherine Forrest: All right, so let me just pause for a second on algorithmic price fixing. And so we know about price fixing, which is just fixing a price, right? Fixing—and we’re going to use the example of two competitors getting together to fix a price. And so algorithmic price fixing is the use of an algorithm by a competitor to fix its price. Now, on its own, there is nothing wrong with using an algorithm, which is really just a mathematical formula, to set a price. In fact, it’s of course, as a matter of law, any company can unilaterally set its price in any manner that it sees fit. The only issues that arise are when two competitors get together and decide to fix price. And so algorithmic price fixing is this new gloss that’s really come into being because an algorithm can take into account many factors like microeconomic factors or macroeconomic factors. And it can be put into a machine learning tool, for instance, to predict what the output might be for a particular year or what the price will be if you look at capacity of a, say, agricultural good at one point in time or not. And so you’ve got a variety of potential inputs in an algorithm. And so it’s nothing more than that. So some kinds of algorithmic price fixing are obviously perfectly legal because there’s just a unilateral fixing by a company of its price.
Scott Caravello: For sure. But so this law is addressed specifically at algorithms that are shared by competitors, that are using information about competitors and disseminating it back to those other companies that then allow them to maintain prices in tandem and not undercut each other. And so whether that may be explicitly or what is often referred to as tacit collusion.
Katherine Forrest: Right. And I think that not undercutting each other is really sort of a key phrase because the concept of a price fix is, hey, if we all fix the price together, then we’ll all be able to keep our prices higher. And that will mean that consumers have to pay higher prices. And that ultimately is one of the evils that the antitrust laws try to prevent, which is higher prices for consumers or reduced output. So tell us more, Scott, about what the bill does—or now actually it’s law because it’s been signed into law.
Scott Caravello: Good point. So for starters, it lowers the pleading standard for the Cartwright claims and also expressly addresses what I mentioned are the so-called common pricing algorithms. So it has both major substantive and procedural implications for the algorithmic and AI pricing landscape.
Katherine Forrest: That’s right. So we know that price fixing between competitors is unlawful under California law under the Cartwright Act and federal law. That’s the Sherman Act that Matthew was talking about. But what AB 325 does is it lowers the pleading standard that a plaintiff has to meet to bring such a claim.
Scott Caravello: Exactly. And so then substantively, it makes it unlawful to use or distribute those common pricing algorithms as part of a contract or combination or conspiracy to restrain trade or commerce.
Katherine Forrest: Right, and so the law is really getting at the use of these algorithms, again, as I’d mentioned before, that some associate with AI and machine learning. And the algorithms can take and synthesize a variety of information collected from a variety of industry participants and can use a formula or an algorithm and set a price. That’s the theory, at least, but I also want to say that for eons, there has been all kinds of public information which industry participants in any number of industries have used to assist them in determining what is the right price. So in some respects, you really want to focus on the confidential aspect of what these so-called pricing algorithms are supposed to be doing. They’re supposed to be combining confidential information because using publicly available information, there’s nothing wrong with that and that’s been done, really, forever. And so the concern is in some ways a concern about a new technology being applied to an old problem. But what we don’t want to do is suggest that something that wasn’t a problem—which is run-of-the-mill individual unilateral fixing of a price using even a sophisticated technology—is somehow suddenly becoming a problem because AI is involved. Unilateral price setting is just fine.
Matthew Robinson: Well, let me jump in here, because I think something very interesting is also included in this bill as it relates to the issue of coercion. So AB 325 makes it unlawful to use or distribute a common pricing algorithm if a person coerces another to adopt recommended prices or commercial terms for the same or similar products or services in California. So what does that mean? It seems what the law is getting at is that one firm can’t use a common pricing algorithm to generate a price and then coerce another firm to adopt the pricing recommended from that common pricing algorithm. This is kind of perplexing because the definition of coercion is actually unexplained in the law. So you can imagine the kind of gap that will be filled from different court practice, from different, from advocacy standpoints as to what constitutes coercion under the law. And that’s, I think, a little newer in the antitrust landscape in terms of what’s actually included in this bill and sort of adding to the antitrust framework over in California.
Katherine Forrest: And remember, it’s not a bill, it’s a law now.
Matthew Robinson: That is correct.
Katherine Forrest: Okay, I think it’s also worth pointing out how AB 325 deals with the definition of person.
Matthew Robinson: Right, so the definition of person under AB 325 continues to exclude end consumers. So this is why that clause on coercion is really focused on firms and companies and not end consumers of the products themselves.
Katherine Forrest: Right. And so let’s pause for a second on the definition of a common pricing algorithm because it’s important that AB 325 defines it functionally as any methodology, including those derived from machine learning or AI, used by two or more persons that uses competitor data to recommend, align, stabilize, set or otherwise influence a price or commercial term.
Scott Caravello: And so notice here what the statute doesn’t do, right? It doesn’t tether itself to whether that competitor data is public or non-public. And I’m just a humble AI lawyer and will defer to the antitrust experts on the line, but that is significant in bringing into scope algorithms that are shared between competitors, even when there’s no confidential data about the competitors’ practices that’s going into the algorithm. So we’re talking about a lot of conduct that could be brought in scope. And so it also doesn’t require that the recommendations that the algorithm put out be binding. The focus is just on the multi-party use, right? So two or more persons using the same or substantially similar methodology and data to shape the price or commercial terms.
Katherine Forrest: Yeah, you know, I’m going to be very interested in some of the cases that are going to start defining how some of this works. And so, you know, I want to clarify another thing that AB 325 doesn’t do. It doesn’t ban or outlaw pricing algorithms used by—again, I’m harping on this—a single firm for its own pricing, right? Because it’s perfectly appropriate for a single firm to use the tools at its disposal to come up with its own price. And it also doesn’t ban or outlaw bespoke tools that a company may have come up with that aren’t “common” per the definition. So in-house or custom tools are perfectly fine.
Scott Caravello: And so if I can, Katherine, I think it could be a good opportunity just to level set a bit further, right? And why are policymakers so concerned about this beyond just the general antitrust concerns that we’ve already discussed? So algorithmic pricing, as you mentioned, is generally an accepted practice. It’s something that we encounter in our daily lives and we have for a long time, right? From purchasing flights to hailing cars on rideshare apps to the prices that we pay on e-commerce platforms. But there is concern, and it’s emerged from the research that looks at algorithmic pricing as driven both by LLMs and by other forms of dynamic pricing technology, that adoption of these tools by competitors could lead to higher than appropriate prices or prices above and beyond what could be sustained in a competitive market. And it’s important to caveat there that a lot of this research has been conducted in simulated settings where the participants may be using different models and algorithms that are learning from one another, which differs from the focus on the shared algorithms at issue here. But we also know, of course, that the real world is much more complicated.
Katherine Forrest: You know, there’s also a reality about the world here that I want to mention that competitors who make similar widgets, say I make one kind of—I’ll go back to my boat example—I make one kind of boat, you make a different kind of boat, my boat does some things, your boat does different kinds of things. But they’re—let’s just imagine for the moment—competitive with one another. They’re subject to some of the same micro and macroeconomic forces, and they might be subject to the same supply chain issues if they’re using similar inputs. You could imagine that they really well could be. And they could have some of the same discounting pressures and they could have some of the same tariff pressures. And so there could be lots of legitimate reasons why pricing between competitors could actually move in parallel, having nothing to do with anything that is in any way illegal. And it could even be that the formulas that they’re using happen to just be picking up on very basic inputs, and that these basic inputs are part of an algorithm that really anybody would use. And so people are going to potentially start finding commonality as if it’s intentionality when that’s not really what’s going on at all. We could have just unilateral behavior.
Matthew Robinson: For sure. And I think this new law, in some ways, is a solution in search of a problem. Because recently what we’ve seen is a number of price fixing cases related to algorithms that have been tossed out. Yes, many of them without prejudice, but tossed out nonetheless. So this is, I think, really an answer to some of the questions and some of the concerns raised by the plaintiffs’ bar.
Scott Caravello: And I think that’s a great point because one of the concerns from an AI perspective is that if the algorithms that industry participants are using become really complicated, maybe it’s going to be very difficult to prove collusion unless the pleading standard is lowered, which—as we mentioned and I think we can get into a bit further—this amendment implements.
Katherine Forrest: Well, one of the concerns I have is that we could end up having a situation, again, from a unilateral perspective, where there might not be any collusion, but if you’ve got a very complicated algorithm, that black box nature of the algorithm could end up seeming to give an inference to conditions of collusion when they really don’t exist. And so I worry about too much regulation in this area.
Matthew Robinson: Right, and it’s really a concern because you have so many different industries that use a variety of algorithms that, for instance, use public data or data that is certainly not sensitive to help them in finding or recommending prices for their products that now could be wrapped up in to this issue.
Scott Caravello: Right, so now that we’ve covered the substantive pillar of the law, why don’t we briefly talk about the pleading reform or what we’ve also referred to as the procedural pillar.
Matthew Robinson: Yeah, Katherine, this is to me very interesting that can be found in AB 325, which is really the lowering of the pleading standard. So typically what you would have to do is ensure that the facts that you’ve alleged account for the possibility that different defendants were actually acting independently. AB 325 removes that requirement and allows you to really just plead the allegations sufficient to show that there is a conspiracy or an agreement, which is certainly new and certainly interesting.
Katherine Forrest: I think it’s—this is going to be a law that’s going to have to be watched. There’s going to be a lot of developments in this area. Scott, we’re going to have to actually see what kinds of algorithms actually get caught up in some of the crosshairs of some of these cases and what the facts are. And then, Matthew, what ends up being considered to be price fixing and what ends up to be just parallel pricing. But that’s our introduction, I think, to this new law, AB 325. Did I get it right, Matthew?
Matthew Robinson: You got it right.
Katherine Forrest: All right. That’s all we’ve got time for today. I’m Katherine Forrest.
Matthew Robinson: I’m Matthew Robinson.
Scott Caravello: And I’m Scott Caravello. Thank you for listening. Don’t forget to like and subscribe.
