Answer capsule
The Justice Department says software does not make shared nonpublic competitor data an independent pricing signal. A CRO needs to know what feeds a pricing tool, what it returns, and who owns the commercial decision.
What the source establishes
- On May 14, 2026, an Acting Deputy Assistant Attorney General described the Antitrust Division's current view of algorithmic conduct in official remarks prepared for delivery.
- The remarks identify software platforms that aggregate competitor data and return pricing recommendations to those competitors as the most developed area of algorithmic antitrust enforcement.
- DOJ says the RealPage consent decree targets nonpublic competitor inputs and granular outputs rather than banning software or algorithmic pricing generally.
- The speaker expressly says the LLM-pricing discussion is not a charging theory or law from the podium; legal treatment depends on the conduct, evidence, agreement, and current authority.
Trace the data before accepting the signal
The direct CRO answer is to stop a pricing deployment when nobody can explain its competitive inputs. Ask which first-party, public, licensed, customer, partner, and competitor data enter the product; whether any input is nonpublic; how current and geographically narrow it is; which users know their information influences other users; and what recommendations, benchmarks, ranges, or alerts come back. A vendor label such as market intelligence, optimization, or AI does not answer those questions.
Keep provenance at field level where it matters. An independently collected public market observation is not the same record as a rival's confidential price, capacity, discount, inventory, bid, or future commercial intention. Derived features can preserve the sensitivity of their inputs even when the dashboard hides company names. If the provider cannot disclose enough for a qualified review, the recommendation should not enter automated price, quote, discount, renewal, or concession decisions.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Keep the commercial decision independent
A pricing recommendation should inform a named decision owner, not replace one. Preserve the business's own current evidence—cost, capacity, inventory, service level, customer value, contract, demand, and approved commercial policy—and require a rationale for the resulting price or concession. Auto-accept, narrow override bands, adherence coaching, and manager pressure can make nominal discretion less meaningful, so inspect how people actually use the output rather than relying on a human-in-the-loop label.
Escalate any design that pools nonpublic competitor information, coordinates users, or makes independent judgment difficult. The appropriate route may be to narrow data, delay it, aggregate it differently, disable a feature, change the approval path, or reject the tool, but that choice needs qualified antitrust review using the actual market and arrangement. This briefing is an operating trigger, not a legal test or safe harbor.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Preserve the configuration and decision record
Inventory every pricing, revenue-management, bidding, quoting, discount, and market-intelligence tool that can influence a commercial term. For each, record the provider, product and model version, configured inputs, training or pooling description, output granularity, automated actions, users, approver, changes, and retention. Keep representative outputs and the final decision with timestamps and reason codes. A later review should be able to distinguish the tool's recommendation from the human decision and the business evidence supporting it.
Monitor changes in data feeds, provider terms, market coverage, model behavior, acceptance rates, override patterns, pricing dispersion, customer complaints, and employee concerns. Give revenue operations, compliance, counsel, procurement, pricing, and product owners a shared escalation path. Preserve useful logs without creating a new warehouse of competitor information or customer data that the business has no reason or right to keep.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Keep the enforcement signal inside its scope
The May 2026 source is an official enforcement speech, not a statute, regulation, final judgment for every pricing product, or decision about the reader's business. It discusses the RealPage consent decree and other cases while emphasizing that courts are sorting different algorithmic arrangements on different records. It also says the forward-looking LLM discussion is not a charging theory or an attempt to make law from a podium.
Use the remarks as a reason to ask the data-flow and decision questions before deployment, then apply current law and qualified advice to the specific competitors, market, communications, vendor terms, model, and conduct. A public-data pricing tool is not condemned merely because it uses an algorithm, and human approval is not a universal cure when the underlying arrangement replaces independent competition. The accountable CRO decision is a bounded, documented authorization—or a stop—based on the real system.
The accountable team should translate this point into a named workflow, affected population, source data, human owner, approval right, exception path, retained evidence, and review date. That translation is what separates an interesting AI development from a decision that can be governed and evaluated.
Decision test
Ask whether the source changes the decision itself, the evidence required, the implementation sequence, or only the language used to describe an existing capability. Record which claims are directly supported, which are provider statements, which require an independent test, and which remain unknown. A source-linked review should make uncertainty easier to see, not bury it inside a blended score.
Questions to take into review
- Which price book and approval matrix apply?
- How are nonstandard terms escalated?
- Which fields may change automatically?
- How are false merges detected and reversed?
- How is error measured across horizons and segments?
- What happens when market conditions shift?
- What evidence defines each stage?
- Which risk factors are causal, correlated, or heuristic?
The publication supports research and executive decision preparation. It does not provide legal, financial, accounting, employment, clinical, cybersecurity, investment, procurement, or implementation advice.