[EDRM Editor’s Note: This article was first published here on May 18, 2026, and EDRM is grateful to Rob Robinson, editor and managing director of EDRM Trusted Partner ComplexDiscovery OÜ, for permission to republish.]

ComplexDiscovery Editor’s Note: Federal antitrust enforcement may be shifting its operational center of gravity. The Justice Department’s Antitrust Division is using artificial intelligence to detect and investigate antitrust violations, according to a May 13 MLex report citing an Antitrust Division official — a development that lands alongside the proposed RealPage consent judgment now in Tunney Act review, sits next to California, New York, and Connecticut’s new state laws, and runs in parallel with the German Federal Cartel Office’s Amazon proceeding.

For cybersecurity, data privacy, regulatory compliance, and eDiscovery professionals, the practical question is shifting from whether the government may read pricing-tool outputs and chat data with machine assistance to how the producing party’s preservation, validation, and production workflows will hold up if it does. The custodial interview reaches the data-science team. The litigation hold reaches the model artifacts. The technology-assisted review protocol has to be documented against a backdrop of symmetric methodology scrutiny.

Watch the next round of state legislative sessions, the conclusion of the RealPage Tunney Act process, and any DOJ disclosure of its AI methodology in a contested production fight. The trajectory is clear; the operational details are what counsel needs to internalize now.

Federal antitrust enforcers are now using artificial intelligence tools in their own investigative work, according to a May 13 MLex report citing an Antitrust Division official. The official said the U.S. Department of Justice is using AI to detect and investigate anticompetitive conduct, with algorithmic pricing and competitor information-sharing among the matters in scope.

The disclosure lands at a moment when the division’s algorithmic-pricing portfolio is already in motion and when state legislatures from Sacramento to Albany have written their own restrictions into law. For companies facing a Hart-Scott-Rodino Second Request, a civil investigative demand, or a grand-jury subpoena, the report turns what had been a theoretical question — does the government read pricing-tool outputs and chat transcripts with machine assistance? — into a working assumption that producing parties have to plan around.

A reported DOJ disclosure follows a clear trajectory

The statement did not come out of nowhere. In November 2025, then-Assistant Attorney General Abigail Slater announced the Justice Department’s proposed settlement with RealPage Inc., the rental-pricing software company whose conduct the division had treated as a test case for algorithmic coordination. Slater said in the announcement that “competing companies must make independent pricing decisions, and with the rise of algorithmic and artificial intelligence tools, we will remain at the forefront of vigorous antitrust enforcement.”

Slater resigned Feb. 12, 2026, weeks before trial in the Live Nation matter. Acting Assistant Attorney General Omeed Assefi has carried the AI line forward. In a May 7 speech at New York University Law School’s Engelberg Center on Innovation Law and Policy, Assefi said the division will continue to test merger-defense narratives that invoke AI as an industry-changing force. “We know you will be tempted to tell us that AI is replacing your industries,” Assefi said. “We get it. We hear that a lot. For us to take it seriously, we expect it to be backed up with actual evidence.” The May 13 MLex report that the division is using AI in antitrust investigations extends that posture from enforcement rhetoric to investigative operations.

RealPage as the operational template

The proposed RealPage consent judgment, filed Nov. 24, 2025, in the U.S. District Court for the Middle District of North Carolina, is the working model that practitioners should study. The proposed seven-year settlement — published in the Federal Register on Jan. 21, 2026, with the court entering the stipulation and proposed order on March 26, 2026, and the response to public comments published May 8 as Tunney Act review continues — would require RealPage to retrain its pricing models on data aged at least 12 months, end real-time use of competitively sensitive nonpublic data, refrain from geographic modeling below the state level, and accept a court-appointed monitor. The settlement terms suggest an investigative record that extended beyond conventional email review and into the data, model, and operational features of pricing systems.

That investigative posture is now publicly reported, with MLex citing an Antitrust Division official as saying the division is using AI in antitrust investigations. If DOJ investigators are applying AI-assisted review or pattern-detection tools to investigative materials, producing parties should assume that communications, pricing records, and system-generated data may receive closer machine-assisted scrutiny than traditional review workflows once suggested. The pattern that has historically been hardest for counsel to explain — parallel pricing behavior tied to shared inputs, decisions made by an algorithm rather than a person, and the trace evidence of an agreement that exists across chat channels and revenue-management dashboards rather than in a single email — is exactly the pattern machine-assisted review is well suited to surface.

States are not waiting, and neither is Berlin

The federal posture coincides with a wave of state and international action. California’s amendments to the Cartwright Act, embodied in AB 325, took effect Jan. 1, 2026, and apply to pricing algorithms across industries. New York amended the Donnelly Act in 2025 to bar algorithmic pricing tools in residential rentals and separately enacted the Algorithmic Pricing Disclosure Act, which requires companies to disclose when an individual consumer’s price has been personalized by an algorithm using that consumer’s data. Connecticut became one of the early states to enact algorithmic-pricing limits when HB 8002 passed in 2025, focused on residential rental markets and on the use of nonpublic competitor information.

By one recent law-firm tally, dozens of bills addressing algorithmic pricing were introduced across over 20 state legislatures during 2025 sessions, with further measures pending in states including Tennessee and New Mexico. State attorneys general are coordinating in parallel, and New York Attorney General Letitia James publicly backed a pair of additional algorithmic-pricing bills in mid-March 2026. The international frame matters too: the German Federal Cartel Office sent Amazon a preliminary legal assessment in June 2025 over algorithmic tools used to set dynamic pricing caps and benchmark third-party seller offers, then prohibited the price-control mechanisms and ordered disgorgement of 59 million euros on Feb. 5 — a parallel proceeding that multinational counsel cannot ignore.

Defense counsel is not conceding the point

Plaintiffs and enforcers have not had a clean run. Through 2025, several algorithmic-pricing cases moved in defendants’ favor at motion-to-dismiss stage, with courts dividing over whether a per se rule or rule-of-reason analysis applies to algorithm-mediated coordination and whether plaintiffs adequately allege the “agreement” that Section 1 of the Sherman Act requires. The defense bar will press that record in cases yet to be filed, and the division’s new acknowledgment that it is using AI in its own investigations opens a parallel front. Producing parties have every incentive to test the methodology of any government model that scored or prioritized their documents — the training set, the validation protocol, the false-positive rate — just as the division tests defense expert work.

What changes inside a Second Request

For eDiscovery teams, the operational shift is concrete. The division’s January 2024 update to its standard preservation specifications already brought ephemeral messaging, Microsoft Teams, Slack, Google Chat, and Signal traffic into scope for HSR Second Requests. Layer AI-assisted government review on top, and the producing party can no longer assume that subtle context — a casual reference in a chat thread, an exception note inside a revenue-management report, a model-tuning parameter buried in an engineering ticket — will go unnoticed.

The producing party’s preservation map has to reach algorithmic-pricing systems themselves. That means the tool’s training data, its model-version history, its configuration files, and its output logs, alongside the human communications that govern its use. Counsel that scopes preservation to email and named custodians will miss the records that the division’s own AI is built to find. For information governance teams, the change has a retention dimension: defensible-deletion schedules that assumed model artifacts and pricing-engine logs were transient operational data need a fresh review against current preservation duty, and lifecycle policies should be aligned with the broader scope of the standard preservation specification.

Preservation, validation, and the documentation that holds up

Practitioners should plan for three workflow changes. First, custodian interviews need to include the data scientists, pricing analysts, and revenue-management staff who configure and operate the algorithms — not only the executives who approve final prices. Second, preservation holds should expressly name pricing-tool databases, model artifacts, and the chat channels where pricing decisions are discussed. Third, technology-assisted review protocols should be documented in a way that anticipates a government request to compare review outputs against the producing party’s own AI tooling.

The validation question runs in both directions. If the government is using AI to score and prioritize documents in a production, the producing party benefits from its own quantitative quality-control measures — sampling, recall and precision metrics, and audit logs — that can withstand scrutiny if a court is asked to weigh the two methodologies. Assefi’s May 7 emphasis on evidence-based advocacy applies internally as well: AI-assisted enforcement is only as defensible as the methodology behind it, and the division will face the same questions from defense counsel that defense counsel will face from the division — what was the training set, how was the model tuned, what is the error rate, and where are the false positives.

For companies and their counsel, the actionable point is direct. The day a Second Request arrives is the wrong day to discover that pricing-tool outputs were not preserved, that the data-science team was not on the litigation hold, or that the production protocol cannot account for AI-driven government review.

How well do your preservation and validation workflows hold up when the regulator on the other side is reading your data with the same class of tools you use to set prices?

Read the original article here.

News sources

• US DOJ using AI to detect anticompetitive conduct, official says (MLex)
• Justice Department Requires RealPage to End the Sharing of Competitively Sensitive Information and Alignment of Pricing Among Competitors (U.S. Department of Justice)
• Acting Assistant Attorney General Omeed A. Assefi Delivers Remarks at Engelberg Center on Innovation Law & Policy at NYU School of Law (U.S. Department of Justice)
• AI Enforcement Accelerates as Federal Policy Stalls and States Step In (Morgan Lewis)
• Algorithmic Pricing Decisions Have Favored Defendants, but the Law Will Continue to Evolve in 2026 (Skadden, Arps, Slate, Meagher & Flom LLP)
• California’s Algorithmic Pricing Antitrust Amendments to the Cartwright Act Take Effect (Morgan Lewis)
• Connecticut Enacts State Antitrust Law on Algorithms (Baker McKenzie)
• Bundeskartellamt has concerns about Amazon’s use of so-called price control mechanisms (Bundeskartellamt)
• United States et al. v. RealPage, Inc. et al. Response to Public Comments (Federal Register (May 8, 2026))

About ComplexDiscovery OÜ

ComplexDiscovery OÜ is an independent digital publication and research organization based in Tallinn, Estonia. ComplexDiscovery covers cybersecurity, data privacy, regulatory compliance, and eDiscovery, with reporting that connects legal and business technology developments—including high-growth startup trends—to international business, policy, and global security dynamics. Focusing on technology and risk issues shaped by cross-border regulation and geopolitical complexity, ComplexDiscovery delivers editorial coverage, original analysis, and curated briefings for a global audience of legal, compliance, security, and technology professionals. Learn more at ComplexDiscovery.com.

Source: ComplexDiscovery OÜ
Assisted by GAI and LLM Technologies per EDRM’s GAI and LLM Policy.