What happened
The U.S. Department of Justice’s criminal antitrust chief told competition lawyers that algorithmic pricing tools and artificial intelligence do not change the basic rule that rivals must make their own pricing decisions, warning that “Software cannot launder collusion.”
Daniel W. Glad, acting deputy assistant attorney general for criminal enforcement in the Antitrust Division, used remarks at the Antitrust West Coast Conference in San Francisco to outline how DOJ is thinking about algorithmic conduct, criminal exposure and the possibility that a large language model could serve as the hub of a pricing arrangement.
Glad pointed to the division’s RealPage resolution as the most developed example of current algorithmic antitrust enforcement. He said DOJ’s November 2025 consent judgment required RealPage to use only historical rental data at least 12 months old, avoid rental-pricing reports narrower than the statewide level, submit to a court-appointed monitor and operate under an antitrust compliance policy approved by the division.
The point of that remedy, Glad said, was not to ban pricing software or algorithmic pricing generally, but to target nonpublic competitor-data inputs and granular outputs returned to competitors. He added that criminal charges remain available when prosecutors can prove competitors used a system to replace independent decision-making with shared competitive intelligence.
Glad also framed the developing case law around a familiar hub-and-spoke question. He contrasted a Seattle federal court’s refusal to dismiss algorithmic price-fixing claims in Duffy v. Yardi with hotel revenue-management cases against Cendyn and Caesars where courts dismissed claims because plaintiffs had not adequately pleaded a rim agreement among competing defendants.
That rim agreement, Glad said, is the hinge for whether a platform arrangement looks like a vertical relationship analyzed under the rule of reason or a horizontal conspiracy potentially subject to the per se rule and criminal enforcement. In the AI context, he said DOJ will ask what a model did with competitors’ nonpublic pricing, capacity or supply data and whether users understood that their inputs would shape outputs delivered to rivals.
The speech also emphasized detection. Glad said the Procurement Collusion Strike Force has trained more than 47,000 federal agents and compliance professionals and produced more than 85 convictions, and he argued that automation can multiply records for investigators: automated bids, competitor-data inputs, language-model pricing queries and parameter changes all may leave a trail.
For companies, Glad said the Antitrust Division’s Whistleblower Rewards Program creates a second race alongside corporate leniency. Instead of only racing other companies to self-report, a company may now be racing employees, former employees, consultants or market participants who have a financial incentive to approach DOJ first.