AI-Driven Drug Pricing Systems and Pharmaceutical Market Control

1. Introduction

AI-driven drug-pricing systems are increasingly capable of analysing enormous datasets involving prescription histories, insurance formularies, competitor prices, reimbursement rules, patient characteristics, demand forecasts, inventory, patent expiry dates, physician behaviour, and pharmacy purchasing patterns. Pharmaceutical manufacturers, wholesalers, pharmacy-benefit managers (PBMs), insurers, and digital health platforms can use such systems to recommend or automatically implement prices.

The competition-law problem arises when AI changes pricing from an individual commercial decision into a coordinated, highly data-driven market-control mechanism.

An AI system may, for example:

• predict the maximum price that a market can bear;
• identify markets with weak therapeutic substitution;
• recommend different prices to different purchasers;
• optimise rebates and formulary placement;
• detect and respond to competitors' prices;
• coordinate prices indirectly through common algorithms;
• exploit information asymmetries between manufacturers, PBMs and patients;
• make switching to competing medicines more difficult;
• reinforce market power following a merger;
• identify opportunities to withdraw or restrict supply;
• facilitate excessive pricing of medicines with few substitutes.

The important legal distinction is that AI itself is not unlawful. Competition-law liability generally depends upon the underlying conduct, market power, agreements, exclusionary strategy, effects, or exploitation.

2. Meaning of AI-Driven Drug Pricing

An AI-driven drug-pricing system may be understood as a technological system that uses machine learning, optimisation, predictive analytics or automated decision-making to determine or recommend:

1. wholesale acquisition prices;
2. pharmacy prices;
3. hospital prices;
4. insurance reimbursement rates;
5. rebates;
6. formulary payments;
7. discounts;
8. geographic prices;
9. patient-specific or payer-specific prices;
10. launch prices for new medicines; or
11. post-patent prices for generic or biosimilar products.

A simplified structure is:

Data → AI model → Demand prediction → Competitive prediction → Price/rebate recommendation → Automated implementation → Market feedback → Model retraining

The final stage is particularly important. An AI system does not merely calculate today's price. It can learn from competitors' reactions and continuously modify future prices.

3. Competition-Law Concerns

A. Algorithmic Price Coordination

One of the most significant risks is algorithmic coordination.

Suppose several pharmaceutical companies independently use sophisticated pricing systems. Each system observes market prices and learns that aggressive price reductions reduce industry profitability.

The algorithms may consequently learn:

"Do not reduce price unless competitors reduce theirs."

No employee may explicitly communicate with another manufacturer. Nevertheless, the market may become substantially less competitive.

The legal question becomes whether this is:

• legitimate independent adaptation;
• conscious parallelism;
• tacit coordination;
• an unlawful agreement;
• facilitation of a cartel; or
• unilateral algorithmic conduct by a dominant firm.

Traditional cartel law generally requires considerably more than merely observing competitors' public prices. But AI makes the evidentiary problem more difficult because coordination can potentially be embedded in software rather than communicated through conventional messages.

4. Excessive Pricing Through AI

AI can also be used to identify the maximum economically extractable price.

For a pharmaceutical product with:

• no close therapeutic substitute,
• high switching costs,
• regulatory barriers,
• strong patent protection,
• limited manufacturing capacity,

an algorithm could estimate the price at which demand remains relatively insensitive.

The system could therefore recommend a very substantial price increase.

This creates potential concerns under excessive-pricing doctrines where applicable.

The European Commission's Aspen matter is particularly relevant. The Commission investigated excessive pricing involving six off-patent cancer medicines and ultimately accepted commitments involving substantial price reductions and supply guarantees.

The AI issue is therefore not simply:

"Did the computer set the price?"

Instead, the legal inquiry can be:

Did a firm possessing market power use an automated system to implement a pricing strategy that would otherwise raise concerns under applicable excessive-pricing or exclusionary-conduct rules?

5. Price Discrimination

AI permits extremely sophisticated segmentation.

A pricing model can distinguish between:

• hospitals;
• government purchasers;
• private insurers;
• PBMs;
• wealthy and low-income geographical markets;
• pharmacies;
• specialty pharmacies;
• different therapeutic indications;
• different levels of purchasing power.

This can produce highly differentiated net prices even where the nominal list price is identical.

Price discrimination is not automatically unlawful. Its legality depends upon the jurisdiction, market position, applicable statute and competitive effects.

The risk becomes greater when discriminatory pricing is used to:

• exclude competitors;
• disadvantage smaller pharmacies;
• prevent parallel trade;
• foreclose generic entrants;
• punish customers for purchasing competing products; or
• exploit a dominant position.

6. AI and Pharmaceutical Market Concentration

AI can increase market concentration because sophisticated pricing systems require:

• enormous datasets;
• computing infrastructure;
• specialised personnel;
• proprietary transaction data;
• historical prescription data;
• payer information;
• pharmacy information;
• real-time competitor information.

Large pharmaceutical firms and pharmaceutical intermediaries may therefore possess a significant informational advantage over smaller rivals.

This can create a data–pricing–market-power feedback loop:

More customers → more data → better AI → better pricing → greater profits → acquisitions → more customers → even more data

Consequently, an apparently ordinary pricing technology may become an important source of durable market power.

7. AI-Enabled Exclusion of Generic Competition

The generic-drug market is particularly sensitive.

An AI system can analyse:

• patent expiration;
• ANDA litigation;
• generic entry probability;
• competitors' manufacturing capacity;
• expected launch dates;
• inventory;
• pharmacy contracts;
• reimbursement levels.

A dominant manufacturer might then optimise its commercial strategy to make generic entry less attractive.

Potential conduct could include:

• strategic discounts;
• exclusionary rebates;
• loyalty arrangements;
• bundling;
• refusal to supply;
• discriminatory access to data;
• manipulation of distribution channels.

The legal problem would not be the predictive algorithm itself but the competitive strategy that the algorithm facilitates.

8. Relevant Case Laws

Case 1: Aspen Pharmacare — Excessive Pricing of Medicines

European Commission, AT.40394, Aspen, 2021

This is one of the most directly relevant pharmaceutical pricing precedents.

The Commission investigated Aspen's pricing of six off-patent cancer medicines. The investigation concerned substantial price increases and the use of supply-related strategies in dealing with national authorities.

The Commission's assessment found that the prices substantially exceeded relevant costs and that Aspen had significant market power in relation to the medicines concerned. The eventual commitments included substantial price reductions and supply obligations.

Relevance to AI

An AI system could make such a pricing strategy substantially more sophisticated by calculating:

• price elasticity;
• substitution possibilities;
• payer resistance;
• withdrawal consequences;
• geographic differences;
• supply vulnerability.

Legal principle: technological sophistication does not immunise a dominant undertaking from scrutiny of potentially excessive pricing.

Case 2: Mylan Pharmaceuticals Inc. v. Warner Chilcott PLC

This litigation concerned competition surrounding pharmaceutical products and allegations involving exclusionary conduct.

The FTC has identified Mylan Pharmaceuticals Inc. v. Warner Chilcott plc among its pharmaceutical competition enforcement matters.

Relevance to AI

The case illustrates why pharmaceutical competition cannot be assessed merely by looking at the final price.

An AI system could be programmed to optimise an entire competitive strategy involving:

• pricing;
• contracting;
• rebates;
• patent-related conduct;
• distribution;
• competitor response.

Therefore, regulators may need to examine the algorithm's objective function and the commercial conduct surrounding it, rather than merely asking whether an individual price is high or low.

Case 3: In re Generic Pharmaceuticals Pricing Antitrust Litigation

U.S. District Court for the Eastern District of Pennsylvania

This extensive litigation concerns allegations of generic-drug manufacturers conspiring to allocate markets and fix prices.

Recent 2026 proceedings continued to address alleged price-fixing involving numerous generic pharmaceutical companies. The litigation includes allegations concerning multiple drugs and an alleged overarching conspiracy.

Relevance to AI

This litigation demonstrates the central distinction between:

Independent pricing intelligence

and

pricing coordination.

If several competitors use systems that merely independently evaluate publicly available market information, that does not automatically establish an unlawful agreement.

But if algorithms are used to implement or facilitate an agreement among competitors, conventional antitrust principles concerning price fixing can become applicable.

AI could make evidence more difficult because the relevant "communication" may exist in:

• model parameters;
• API calls;
• shared datasets;
• software instructions;
• automated bidding systems;
• common pricing platforms.

Case 4: FTC v. Surescripts LLC

U.S. District Court for the District of Columbia

Surescripts concerned electronic prescription infrastructure rather than pharmaceutical manufacturing prices directly.

The FTC alleged that Surescripts maintained monopolies in electronic-prescription routing and eligibility markets through vertical and horizontal restraints, including restrictions that discouraged participants from using competing platforms.

Relevance to AI

The case is significant because modern pharmaceutical pricing increasingly depends upon digital infrastructure.

An AI pricing platform can become strategically important if it controls:

• prescription data;
• insurance eligibility information;
• pharmacy data;
• reimbursement information;
• pricing interfaces.

Thus, competition authorities may have to analyse not only the price but also the digital infrastructure through which the price is generated and transmitted.

9. Case 5: In re Caremark Rx / Zinc Health Services — Insulin

FTC, 2022–2026

This is highly relevant to modern drug-pricing systems.

The FTC alleged that major PBMs and affiliated group-purchasing organisations used rebate practices that artificially increased insulin list prices and affected competition based on rebate levels rather than net prices.

In 2026, settlements with Express Scripts and Caremark addressed these allegations and required significant changes to rebate and pricing practices.

Relevance to AI

This demonstrates why an AI pricing investigation must distinguish:

List price

from

net price

from

rebate-adjusted price

from

patient out-of-pocket price.

An AI system optimising rebates could theoretically produce a situation in which:

Manufacturer → high list price → large rebate → preferred formulary position → intermediary revenue → higher patient exposure to list price.

Consequently, an AI system could optimise an economically complex structure even when the nominal price alone does not reveal the competitive effect.

10. Case 6: Aurobindo Pharma–Lannett

FTC, 2026

The FTC required Aurobindo Pharma to divest four generic drug products as a condition of completing its acquisition of Lannett. The FTC stated that the transaction raised concerns relating to competition and potential higher drug costs.

Relevance to AI

This illustrates the interaction between AI, mergers and pharmaceutical concentration.

Suppose a merger combines two major datasets containing:

• prescription demand;
• competitor pricing;
• hospital purchasing;
• generic entry data;
• reimbursement information.

The merged firm may obtain a significantly stronger pricing-information advantage.

Therefore, merger analysis may need to consider not merely traditional market shares but also:

• data concentration;
• algorithmic capabilities;
• pricing infrastructure;
• access to commercially sensitive information;
• AI-driven entry barriers.

11. Case 7: United States v. Microsoft — Algorithmic/Platform Analogy

Although not a pharmaceutical case, the Microsoft monopolisation litigation provides a useful analytical analogy.

The case demonstrates that competition law can address conduct through which a firm with substantial market power uses control over an important technological platform to restrict competitive opportunities.

Relevance to pharmaceutical AI

A pharmaceutical pricing platform could become analogous to a strategic platform if it controls:

• pricing data;
• pharmacy access;
• formulary information;
• prescribing information;
• competitor information;
• automated contracting.

The underlying principle is that technology does not place exclusionary conduct outside competition law.

12. Case 8: AKZO Chemie BV v Commission

Court of Justice of the European Communities

AKZO is an important predatory-pricing precedent.

The case established important principles concerning assessment of prices below relevant cost benchmarks and exclusionary pricing.

AI relevance

An AI system can make predatory pricing more sophisticated.

For example, it could calculate:

• competitor-specific vulnerability;
• duration of below-cost pricing;
• expected exit probability;
• switching behaviour;
• geographic targeting;
• post-exit recoupment opportunities.

An algorithm could therefore transform predatory pricing from a simple uniform strategy into a highly targeted strategy.

The legal question would remain whether the underlying pricing conduct satisfies the applicable legal test.

13. AI and Predatory Pricing

A hypothetical AI-driven predatory strategy could look like this:

Stage 1: AI identifies a financially vulnerable generic competitor.

Stage 2: It calculates the competitor's estimated break-even price.

Stage 3: It recommends prices below that level in the competitor's important markets.

Stage 4: It maintains higher prices elsewhere.

Stage 5: After competitor withdrawal, prices are increased.

The competitive concern is therefore not merely "low prices."

The concern is whether low prices are being used as part of an exclusionary strategy.

14. AI-Driven Rebate Optimisation

Pharmaceutical pricing frequently involves rebates.

An AI model could optimise:

Net Price=List Price−Rebate−DiscountNet\ Price = List\ Price - Rebate - Discount

But a more sophisticated model could simultaneously optimise:

Profit=Revenue−Rebates−Distribution Costs−Acquisition CostsProfit = Revenue - Rebates - Distribution\ Costs - Acquisition\ Costs

subject to:

• formulary placement;
• competitor prices;
• market shares;
• payer preferences;
• regulatory restrictions.

This can produce very complex pricing structures.

The Caremark/insulin proceedings illustrate why competition authorities may need to examine the entire pricing architecture rather than focusing only on advertised list prices.

15. Algorithmic Discrimination Between Buyers

AI could classify buyers into categories:

Such differentiation can have legitimate commercial explanations.

However, competition concerns can arise where discrimination is used by a dominant undertaking to:

• foreclose competitors;
• disadvantage independent distributors;
• prevent switching;
• restrict parallel trade;
• punish customers that deal with rivals.

16. AI and Market Foreclosure

A pharmaceutical AI system may optimise not merely price but market access.

It could rank customers according to their importance and determine:

Which hospitals should receive discounts?

Which pharmacies should receive favourable supply terms?

Which distributors should receive inventory?

Which customers should receive loyalty rebates?

Such optimisation can potentially reinforce an incumbent's market position.

Therefore:

Pricing algorithm + distribution algorithm + contracting algorithm

may collectively create greater competitive effects than any one algorithm considered separately.

17. AI and Pharmaceutical Mergers

AI also affects merger control.

A pharmaceutical merger can combine:

• proprietary clinical data;
• pricing data;
• prescription data;
• patient demand information;
• manufacturing data;
• competitor intelligence.

The merged company may therefore acquire a much stronger ability to predict market behaviour.

Potential competition concerns include:

1. elimination of a competing pricing model;
2. concentration of pricing data;
3. reduction of independent price discovery;
4. increased ability to personalise prices;
5. increased bargaining power against PBMs;
6. enhanced ability to identify and suppress generic entry.

The FTC's 2026 Aurobindo-Lannett action illustrates the continuing significance of maintaining competitive alternatives in generic-drug markets.

18. AI and Pharmaceutical Market Control

The concept of market control is broader than price fixing.

AI can provide control over:

A. Price

Determining the amount charged.

B. Quantity

Determining how much medicine is supplied.

C. Access

Determining which purchasers receive the product.

D. Timing

Determining when discounts or price increases occur.

E. Information

Determining which market participants receive commercially important data.

F. Switching

Determining whether customers receive incentives to remain with the incumbent.

G. Entry

Predicting and responding to potential generic or biosimilar entrants.

Thus, the competitive concern can be represented as:

AI pricing → AI contracting → AI distribution → AI information control → stronger market power

19. Evidence and Discovery Problems

AI creates an important evidentiary challenge.

Traditional antitrust investigations may examine:

• emails;
• contracts;
• telephone records;
• meeting minutes;
• spreadsheets.

AI-driven pricing may instead involve:

• source code;
• model weights;
• prompts;
• training datasets;
• API logs;
• optimisation objectives;
• model outputs;
• automated pricing histories;
• audit trails;
• reinforcement-learning records.

Therefore, competition authorities may need to determine:

What exactly did the algorithm know when it made the pricing decision?

and:

Who designed the objective that the algorithm was optimising?

20. Human Responsibility for AI Decisions

A company cannot necessarily avoid competition-law scrutiny merely by claiming:

"The AI made the decision."

Legal responsibility generally turns upon the conduct attributable to the undertaking and applicable statutory principles.

Important questions include:

1. Who designed the algorithm?
2. What objective was programmed?
3. What data was supplied?
4. Was competitor-sensitive information used?
5. Was the model instructed to follow competitors?
6. Were employees monitoring its outputs?
7. Were unlawful outcomes foreseeable?
8. Did management approve the pricing strategy?
9. Did the company intervene when problematic behaviour became apparent?
10. Was the AI deployed to implement an existing anticompetitive agreement?

21. Competition-Law Tests

A. Agreement or Concerted Practice

Authorities may investigate whether firms:

• exchanged commercially sensitive information;
• agreed on pricing parameters;
• used a common pricing platform;
• coordinated algorithms;
• agreed to follow algorithmically generated prices.

The central question is whether there is sufficient evidence of coordinated conduct under the relevant jurisdiction.

B. Abuse of Dominance

Where a pharmaceutical undertaking is dominant, authorities may investigate:

• excessive pricing;
• predatory pricing;
• discriminatory pricing;
• loyalty rebates;
• tying;
• bundling;
• refusal to supply;
• exclusionary contracting.

The use of AI can be evidence of how the conduct was implemented but does not automatically create liability.

C. Merger Control

Authorities may consider whether a transaction:

• removes an important competitor;
• concentrates critical datasets;
• strengthens pricing power;
• increases algorithmic capabilities;
• creates barriers to generic or biosimilar entry.

D. Data-Related Competition Concerns

A pricing algorithm can become competitively significant where the underlying data is:

• difficult to replicate;
• commercially sensitive;
• continuously updated;
• unavailable to competitors;
• necessary for accurate pricing.

22. Pharmaceutical-Specific Risk Matrix

23. Defences and Legitimate Uses of AI

AI pricing is not inherently anticompetitive.

Legitimate uses may include:

• reducing inventory waste;
• forecasting shortages;
• improving supply reliability;
• detecting counterfeit medicines;
• reducing distribution costs;
• forecasting manufacturing requirements;
• ensuring regulatory compliance;
• improving procurement efficiency;
• reducing transaction costs.

Competition law should therefore distinguish efficient AI optimisation from AI-assisted restriction of competition.

The European Commission itself recognises that pharmaceutical markets involve lengthy, risky and expensive innovation processes and that competition analysis must account for both affordability and innovation.

24. Compliance Framework for Pharmaceutical AI

Pharmaceutical companies deploying AI pricing systems should maintain:

1. Independent pricing controls

AI should not automatically follow competitor prices without appropriate legal review.

2. Data governance

Competitor-sensitive information should be carefully controlled.

3. Audit trails

The company should preserve:

• inputs;
• outputs;
• model versions;
• pricing instructions;
• human interventions.

4. Competition-law testing

Models should be tested for outcomes involving:

• coordinated pricing;
• exclusionary discounts;
• discriminatory access;
• predatory pricing.

5. Human oversight

High-impact pricing decisions should have appropriate human review.

6. Model documentation

The company should document:

• objective function;
• training data;
• constraints;
• pricing parameters;
• decision rules.

7. Merger controls

Acquisition of large pharmaceutical datasets should be considered in competition assessments.

25. Six Core Legal Lessons

The case law collectively supports several important propositions:

First, AI does not create a separate immunity from competition law.

Second, pharmaceutical pricing must sometimes be examined together with supply, contracting and distribution strategies.

Third, dominant-firm pricing can raise concerns even where there is no conventional cartel.

Fourth, digital infrastructure can itself become a source of market power, as illustrated by Surescripts.

Fifth, pharmaceutical intermediary practices can materially affect effective drug prices, as demonstrated by the insulin/PBM proceedings.

Sixth, concentration of pharmaceutical products and data can create competition concerns at the merger stage, illustrated by the Aurobindo-Lannett matter.

26. Conclusion

AI-driven drug pricing systems represent a transition from conventional pricing to continuous computational market optimisation. The competition-law significance lies not in the mere use of artificial intelligence, but in what the system is optimising, what information it uses, whether it facilitates coordination, and whether it strengthens or exploits market power.

The most important legal distinction is:

AI-assisted independent pricing is not equivalent to unlawful price coordination.

Nevertheless, an AI system can potentially magnify traditional competition-law problems by making pricing:

• faster,
• more precise,
• less transparent,
• highly personalised,
• responsive to competitors,
• integrated with rebates and distribution,
• and capable of continuous optimisation.

The Aspen decision demonstrates the relevance of excessive-pricing analysis in pharmaceutical markets; the generic-drug pricing litigation demonstrates the continuing importance of price-fixing principles; Surescripts illustrates the importance of digital infrastructure; the insulin/PBM proceedings demonstrate the complexity of rebate-driven drug pricing; and the Aurobindo-Lannett merger action illustrates the continuing importance of maintaining competitive alternatives in generic medicines.

Accordingly, the emerging legal framework can be expressed as:

AI Pricing Technology → Data Advantage → Pricing Optimisation → Contracting/Distribution Effects → Market Power → Potential Coordination or Foreclosure → Competition-Law Review

The central regulatory challenge is therefore to