Competition regulators are grappling with the role of artificial intelligence in relation to competitive intelligence — with particular attention to the grey area between lawful parallel conduct and illegal price-fixing.
Competition regulators are grappling with the role of artificial intelligence in relation to competitive intelligence — with particular attention to the grey area between lawful parallel conduct and illegal price-fixing. The classic illustration of the distinction involves two gas stations across the street from each other. When those stations raise their prices to the same level at nearly the same time, they are running afoul of the Sherman Act only if they are actively colluding, rather than passively observing each other.
The adoption by companies of artificial intelligence to analyze competitors raises new questions about active and passive behavior. Participants at a conference* Monday wrestled with the regulatory implications of AI — and generally acknowledged that the antitrust world is in the early days of assessing the technology.
University of Chicago law professor Eric Posner of MoloLamken LLP cited the gas station narrative. "If the other gas station only keeps the price high, I keep the price high. If they lower it, I lower mine. ... This is what economists call ‘tit-for-tat.’ "
Researchers, he said, have found that "very neutrally designed algorithms, algorithms which are told simply to maximize revenue, not to coordinate prices, will hit upon a very similar strategy — kind of a different kind of strategy, but they do it much better and much faster than humans do.”
Artificial intelligence, Posner said, can produce collusive effects even if competitors aren’t working together.
“The problem here is that, unlike in the shared-algorithm setting or shared-data setting, the firms can be acting in a very independent way, they could be using different algorithms, but nonetheless, they could end up with pricing that effectively is collusive."
Posner isn't convinced that traditional antitrust law is able to deal with this issue.
"I think you would have to change a great deal," he said. "The traditional reason why courts decline to punish companies that engage in that kind of ‘follow the leader’ or coordinated conduct ... is the problem of a remedy. It's very difficult to say to someone, ‘You have to pick a price without paying attention to any of the information out there.’ "
If competitors’ prices are available to AI on the web, it is difficult to imagine how antitrust regulators could prohibit algorithms from referencing the information, Posner said.
“That is the problem,” said Elizabeth Odette, an assistant attorney general in Minnesota and the chair of the National Association of Attorneys General Antitrust Task Force, “because I see some draft legislation that doesn't go after a specific industry and may not even define AI or algorithmic pricing appropriately.”
Vague, inadequate guidance puts enforcers in a difficult position, Odette said.
“I think where we've seen success is by focusing on an industry,” she said. “So there's been a number [of cases], in cities and states that focus on rent or other things … You'll probably have more success with states and cities focusing on specific industries where they're identifying a specific problem.”
Clarity is important, she said. “It may be not impossible, but very difficult, for state enforcers to be figuring out what is a violation if it's not defined.”
Daniel Graulich, chief of staff and attorney advisor to US Federal Trade Commissioner Mark Meador, said the Department of Justice is preparing guidance on collaboration among competitors (see here).
“I think any of these questions are fair game for the competitor collaboration guidelines,” Graulich said.
But he also said traditional guidance remains useful.
“If you’re talking about a technology that’s individualized, customized, providing you [an] individualized solution not really based on your competitors, that’s maybe your low-risk bucket,” Graulich said. “If it’s a tool that’s aggregating … insights across companies that compete with each other, that tends to be where you get into more antitrust concerns.”
Graulich said enforcers should examine the market coverage of the tool, whether competitively sensitive information is exchanged, any limits on information sharing, and whether a company is getting access only to its own inputs.
Members of the panel considered the possibility that separate algorithms can formulate price recommendations, or even coordinate, resulting in the charge of supra-competitive prices without the involvement of competitors or sensitive information.
“It’s possible we get to this situation where there are multiple different algorithms. They’re not designed to collude … They learn to collude and talk to each other, and that starts to become a sticky situation for antitrust law and what is an agreement,” said Adam Cella, chief counsel for the US House Judiciary Subcommittee for Administrative State, Regulatory Reform and Antitrust.
Enforcers, he said, are empowered to bring cases when consumers are harmed. Companies’ self-interest will ensure they will know when “their tools are colluding.”
Ashley Walters, assistant attorney general in the District of Columbia, said that AI can, in effect, function as the hub in a hub-and-spoke pricing conspiracy where information comes inappropriately from a central source — in this case, the artificial intelligence.
“AI is really in a risky place here with different kinds of vertical and horizontal agreements where a hub can coordinate … whether intentionally or not, anti competitive dynamics,” she said. The spokes, she said, “essentially can form agreements without any direct communication.”
These ambiguities create a problem for enforcement, Posner said.
“I suspect it’s not going to be easy to distinguish [between] algorithms that are just used by companies to set prices in a reasonable way and algorithms that are used that set prices by taking into account the prices of other firms,” Posner said. “I think we’re going to have to see first how aggressively firms use these algorithms, how effectively they are [used] in the real world, before we’ll know what to do about it.”
*"Washington Antitrust and Digital Markets Forum," MLex, George Washington University Competition Law Center, Forum Global, Washington, DC; March 23, 2026.
Please email editors@mlex.com to contact the editorial staff regarding this story, or to submit the names of lawyers and advisers.