Algorithmic Antitrust Risk: The Newest AI Compliance Challenge

Antitrust enforcers at the Department of Justice have begun warning companies about a risk they call "algorithmic antitrust" — the possibility that an AI pricing tool could trigger the same price-fixing, bid-rigging, or market-allocation violations that have been illegal since the Sherman Act of 1890. The danger arises not from any single company using its own data to set prices, but from situations where multiple competitors feed confidential, nonpublic pricing information into a shared algorithm that then spits out coordinated recommendations. This webinar, hosted by Nick Gallo of Ethicsverse with radical compliance commentator Matt Kelly and antitrust counsel Dan Shulak of Hogan Lovells Cadwalader, unpacks what the misconduct is (and isn't), where enforcement is heading, and what compliance officers should actually do about a risk that lands squarely in the gray area between IT, sales, and legal. The panel also examines the DOJ's leniency program and its newer whistleblower awards, which together create a high-stakes race to self-report — a race an outside individual whistleblower can now win before the company even knows it has a problem.

The central thesis advanced by the panel is one of legal continuity rather than novelty: the hard-core antitrust prohibitions against price-fixing, bid-rigging, and market allocation codified in the Sherman Act apply with equal force whether coordination is achieved through a smoke-filled room, an encrypted messaging application, or a shared pricing algorithm. The discussion delineates the analytical framework enforcers are applying — attending to what data enters a tool (proprietary versus co-mingled competitor information), how recommendations are generated (public versus nonpublic inputs), and whether outputs are binding or subject to independent human judgment. The panel situates this within an evolving enforcement landscape characterized by DOJ speeches (notably those of Acting Deputy Assistant Attorney General Daniel Glad), ongoing but as-yet-uncharged criminal investigations, concluded civil matters such as RealPage and MultiPlan, and a proliferating patchwork of subnational regulation, exemplified by California's 2026 amendment to the Cartwright Act.

Key Takeaways

The Same Old Laws Apply to a Brand-New Technology

• The conduct at the heart of algorithmic antitrust is not a new category of offense but the same hard-core price-fixing, bid-rigging, and market-allocation prohibition that has existed since the Sherman Act of 1890.

• Enforcers have signaled a consistent through-line dating back to the Biden administration: agreements among competitors remain fair game for criminal prosecution regardless of whether they are facilitated by an algorithm or by more traditional means.

• The governing principle is intuitive — if an outcome would be illegal when arranged over a WhatsApp thread or a conference call, it does not become legal simply because it was produced on a server or by a large language model.

The Data Determines the Risk

• Independent use of an AI pricing tool that draws only on a company's own data — without competitor involvement — is generally considered acceptable and does not, on its own, create antitrust exposure.

• The danger zone opens when multiple competitors feed their confidential, nonpublic pricing information into a single shared algorithm that then generates recommendations based on that co-mingled competitive data.

• Compliance officers should interrogate three questions for any pricing tool: whose data goes in, what the recommendation is based on, and what the organization is contractually or operationally obligated to do with the output.

Enforcement Is Real, Even Without a Criminal Case Yet

• As of the webinar, the DOJ's Antitrust Division had not yet brought a criminal enforcement action based on an algorithmic pricing tool, but panelists confirmed that investigations are active and ongoing.

• Concluded matters such as RealPage in residential real estate and MultiPlan in healthcare demonstrate that civil enforcement and private litigation are already reshaping the risk landscape.

• The absence of a criminal conviction to date should be read as an early-stage warning rather than reassurance, because enforcers have publicly telegraphed their intent through speeches and resource commitments.

Certain Industries Face Heightened Scrutiny

• Enforcement activity has clustered in real estate, hospitality and hotels, and healthcare, making organizations in those sectors especially exposed to algorithmic antitrust review.

• Government contracting represents a particularly high-risk arena because standardized bidding, documented communications, repeat interactions, and institutional record-keeping make collusion far more detectable when layered with AI-driven bid tools.

• The DOJ's Procurement Collusion Strike Force has trained tens of thousands of agents and prosecutors to detect public procurement fraud, meaning companies engaged in government bidding operate in an intensely monitored environment.

Regulation Is Multiplying Beyond the Federal Level

• California moved to the front of state enforcement by amending its Cartwright Act to ban agreements among competitors to use a common pricing algorithm that relies on competitor data, a law that took effect January 1, 2026.

• The California statute contains a critical carve-out: it does not apply when a business uses an algorithmic pricing tool drawing solely on the business's own data, reinforcing the centrality of the data-input question.

• Enforcement is expanding internationally as well, with sophisticated competition authorities in the EU, Brazil, Australia, and Canada all prioritizing algorithmic pricing cases, while several U.S. municipalities have introduced housing-specific legislation.

AI May Make Collusion More Visible, Not Less

• A traditional price-fixing conspiracy conducted by a handful of executives at a cocktail bar involves only those few individuals, but an AI-mediated scheme can implicate the vendor, the data scientist who trained the model, and the IT team.

• This expanded surface area means that far more people may become aware of a potential conspiracy, increasing the likelihood that someone reports it internally or externally.

• The panel noted the irony that hiding misconduct "behind the AI" may actually generate a larger trail of digital evidence and a wider pool of potential whistleblowers than old-fashioned collusion ever did.

No One Clearly Owns This Risk

• A listener's question exposed the core governance problem: the tool is a technology matter, but the IT team isn't the user, sales is — yet if sales misuses it, the fallout becomes a compliance problem.

• This ambiguity creates a structural blind spot in which no single function is accountable for evaluating the antitrust implications of an AI pricing tool before it is deployed.

• The panelists argued that closing these "open loops" ultimately falls to compliance, which must insert itself into vendor evaluation and AI governance conversations rather than waiting for the risk to materialize.

Traditional Policies Are Not Enough

• Existing antitrust compliance policies typically reference the smoke-filled room and, more recently, texting and ephemeral messaging apps, but they rarely address AI-specific scenarios.

• The panel recommended that organizations adopt an AI-specific antitrust policy — either as a standalone document or a dedicated subsection of an existing policy — as the essential starting point for managing this risk.

• Because the DOJ evaluates whether compliance programs are tailored to an organization's actual risks, companies using these tools are expected to adapt their policies, training, and messaging to reflect that specific exposure.

Risk Assessments and Vendor Diligence Are Practical First Steps

• The panel advocated conducting an AI pricing tool risk assessment for existing tools and mandating a formal review before any new tool is deployed, examining whether it uses competitor data, whether that data is public or nonpublic, and how recommendations are generated.

• Particular scrutiny should be applied to shared third-party platforms used across an entire industry, as illustrated by MultiPlan, where nearly every major health insurer relied on the same system.

• Where feasible, compliance teams should seek contractual representations around data segregation from vendors, since the inability to secure such assurances signals a materially riskier arrangement.

The Leniency Race Makes Speed a Strategic Imperative

• The DOJ's leniency program offers full immunity from criminal prosecution to the first cartel participant to self-report, creating a prisoner's-dilemma dynamic in which being first is the decisive advantage.

• The newer whistleblower award structure — which has already produced one deliberately headline-grabbing $1 million payment — introduces a second race, because an individual whistleblower who reports first can eliminate the company's ability to self-disclose.

• Under the 2022 promptness requirement, a compliance officer's awareness of a potential violation starts the clock on self-reporting, meaning concerns must be escalated to legal and senior leadership immediately rather than sat upon while an internal investigation drags on.

Closing Summary

Algorithmic antitrust is best understood not as a new prohibition but as an old one wearing unfamiliar clothing — the Sherman Act's ban on coordinated pricing, reasserted for an era in which competitors may collude through a shared algorithm rather than a phone call. The through-line of this discussion is that the analytical questions remain stubbornly consistent: what data goes in, what the output is based on, and what the organization does with it. What has changed is the difficulty of seeing the risk at all, given that responsibility is diffused across sales, IT, legal, and outside vendors, and that most organizations lack even a basic inventory of their AI use cases. For compliance and ethics professionals, the webinar's imperatives are clear and actionable: build that inventory, draft AI-specific antitrust policies, embed compliance in vendor evaluation and pre-deployment review, and internalize the high-stakes urgency created by the leniency and whistleblower programs, where the first to report wins and a compliance officer's own awareness starts the clock. In a domain still defined by legal gray areas and evolving enforcement, the organizations best positioned to avoid a headline are those treating algorithmic antitrust as a present-tense governance priority rather than a hypothetical future problem.