Can Conspiracies to Better the World Be Anticompetitive?
“[R]egulators may not care what the ostensible goal of a conspiracy is if they think it might harm competition. Indeed, the law generally does not recognize health and safety as justifications for anticompetitive conduct.”
The Federal Trade Commission (FTC) screwed up. At least that’s how it explains what it calls the “lock[ed] in exploitative business models and monopoly power” of today’s internet giants. It blames “delayed government action.” But the agency says it won’t allow the same thing to happen with artificial intelligence (AI). With AI, the FTC “plan[s] on using the full scope of [its] authority to make sure that history does not repeat itself.”
In response, AI developers … agree? Sort of. Sam Altman – CEO of OpenAI – famously (or infamously) said, “AI will probably most likely lead to the end of the world, but in the meantime, there’ll be great companies.” Now, he wants those companies to work together to avoid the apocalypse by sharing best practices. He has specifically mentioned that “everyone is very careful not to run afoul of antitrust law, in coordinating with competitors.” Indeed, coordinating technological development necessarily limits the capabilities and potential usefulness of AI systems, which raises the question – can a conspiracy to save the world from AI domination nonetheless be anticompetitive?
These issues impact antitrust practice and policy, not just in AI and Tech, but industries across the economy.
A Conspiracy is a Conspiracy is a Conspiracy. Or is it?
Let’s talk briefly about another potentially world-threatening issue for a moment: climate change. In 2019, the Department of Justice (DOJ) opened an antitrust inquiry into four of the world’s largest automakers after they struck a deal with California to reduce automobile emissions. While not everyone may agree on the necessity of emission reduction, it’s at least intended to help the world, right? But announce a deal with your competitors to combat climate change, face an antitrust investigation. The DOJ’s rationale? The deal could potentially limit consumer choice. The investigation was dropped fairly quickly, but it shows regulators may not care what the ostensible goal of a conspiracy is if they think it might harm competition. Indeed, the law generally does not recognize health and safety as justifications for anticompetitive conduct.
More recently, and across the pond, the EU last December issued antitrust guidance for certain sustainability agreements between competitors. FTC Chair Lina Khan, in contrast, made it clear in remarks earlier this year that there is no antitrust exemption for environmental issues. She said the same thing about AI: “There is no AI exemption to the laws on the books.” Indeed, the FTC is in the process of probing AI strategic partnerships – it issued compulsory information requests to the big AI companies to identify their “investments in or partnerships with” others, and for each, to “explain each strategic rationale for the transaction….”
But this is not just a Sam Altman problem. The FTC and DOJ have taken a hard look at information exchanges among competitors generally, rescinding the longstanding Antitrust Enforcement Policy in Health Care, which competitors in many industries, not just healthcare, relied on for its so-called “safe harbors” of information sharing, when information was sufficiently old, anonymized, and aggregated.
So, what to do if competitors want to get together to better the world? Mr. Altman’s suggestion to have a “lawyer in the room” is not a bad one, but it also will not save an otherwise anticompetitive agreement. First and foremost, of course, agreements should be avoided if at all possible. Sharing best practices, on the other hand, can be legitimate and procompetitive if it does not veer into coordination of prices, production, or the like. If an actual agreement is required to effectuate the world-benefiting goal, parties may consider the feasibility of establishing a bona fide joint venture, which can take advantage of the single-firm rule established by the Supreme Court. Under that rule, the joint venture will be subject – not to per se condemnation like price-fixing conspiracies – but to the rule of reason, under which the parties may be able to argue activities designed to have a significant impact on health or safety should not be deemed a “naked”, per se restraint on trade but a procompetitive one.
Collecting and Sharing Information in the AI Age
Information sharing among competing AI companies may be one thing, but what about using AI to share or obtain data? Data has never been more prevalent, or more usable. Algorithms can ingest and analyze massive quantities of data. The DOJ and FTC are concerned that competitors can use these tools to glean information about competitors that otherwise would have remained hidden in a mound of ones and zeros. But if the data is public, there should, in theory, be no concern. Competitors are always on the lookout for competitive intelligence, and if an algorithm refines and deepens that capability, that should be seen as competition enhancing.
Let’s take it one more step. What if algorithms are used not just to collect and analyze reams of publicly available data, but to make competitive decisions based on the data – say, setting prices. Online marketplaces are full of algorithmically set prices, and economists have begun to study the results. Assume, for purposes of this discussion, that the algorithms are put into place to better the world: to lower prices more often. Good for consumers, right? Not everyone thinks so. The FTC and DOJ are reading studies like this one published by the Brookings Institute — finding that “even very simple pricing algorithms can raise prices.”
But there’s also widespread “agree[ment] that algorithms that adjust prices based on demand conditions and/or costs have the potential to increase efficiency.” And the faster the algorithm, the “lower [the] prices.” But to make the point that simple algorithms can nonetheless lead to higher prices, the study’s authors use an example of a slow, traditional pricing retailer and a fast, algorithmically pricing retailer:
“Consider a retailer that programs a pricing rule that quickly undercuts a rival’s price by $3. Because the rule is enforced by a computer, the retailer can tie its hands to this strategy, regardless of the price chosen by its rival. For this example, suppose the competitive price level without algorithms is $15. One might intuitively think that a commitment to undercut a rival’s price would lead to lower prices. However, economic theory says the opposite. Why?”
The driving goal of traditional economic theory is to find the sweet spot: the profit maximizing price. Normally, one would think that if the faster rival lowers price, the slower retailer will lower price in return. But if the slower retailer knows or realizes that the faster rival is pricing based on an algorithmic rule, its ability to gain share by further cutting price is cut off. So, its profit maximizing strategy shifts and will, according to the authors, typically “lead to prices that are higher than the [otherwise] competitive price levels for both firms….”
But this is just one scenario with numerous assumptions built in to make a point. So, the authors imagined a world where “all competitors had high-frequency, autonomous algorithms.” What then? Higher prices. And, according to some (admittedly “unsophisticated”) simulations, even algorithm-to-algorithm collusion. But the studies also show that such algorithmic collusion can be relatively easily avoided by telling the algorithm some basic rules, such as lower prices lead to higher shares.
As to the possibility that the spread of algorithmic pricing may lead to higher prices even without any collusion, whether human or algorithmic, well, the FTC and DOJ have a problem: the law criminalizes anticompetitive agreements, not unilateral activities that happen to increase prices. But where algorithm-users should be careful, and conduct their due diligence, is in determining what information the algorithm uses in setting prices. Is it all legitimate, publicly available data, or does it include non-public competitively sensitive information from competitors?
Also take care when an industry is dominated by a single information service provider, including price recommendation services. Those are the allegations in hotel algorithmic-price fixing cases targeting Vegas and Atlantic City hotels litigation. (Gibson et al. v. MGM Resorts Int’l et al., No. 2:23-cv-00140-MMD-DJA (D. Nev.); Cornish-Adebiyi et al. v. Caesars Entertainment, Inc., No. 1:23-cv-02536-KMW-EAP (D.N.J.); Blair-Smith v. Caesars Entertainment, Inc. et al., No. 1:23-cv-06506-KMW-EAP (D.N.J.); Fabel v. Boardwalk 1000, LLC et al., No. 1:23-cv-06576-KMW-EAP (D.N.J.).
The Vegas complaints have been twice dismissed now, largely because the plaintiffs could not allege that the algorithm used anything other than a hotel’s own, and public, information, and could not allege an agreement among hotels to abide by the algorithm’s recommendations. The court there noted that, “If they all agreed to outsource their pricing decisions to a third party, and all agreed to price according to the recommendations provided by that third party, it would be plausible to infer the existence of a collusive agreement to fix prices.” So, don’t do that.
Image Source: Deposit Photos
Author: md3d
Image ID: 182090646
