Airlines are racing to replace rigid fare structures with real-time, data-driven pricing. The shift promises sharper offers, higher revenue and a better match between what passengers want and what carriers can sell. Yet the same technology that unlocks this value carries a quiet but serious legal danger: algorithmic systems can produce collusive-like outcomes without any human ever agreeing to fix prices.
This article explains the risk in plain terms, shows exactly when and why it crosses into unlawfulness, and sets out what airlines can and should do—both to stay compliant and to capture the legitimate commercial upside.
The Conceptual Process: How Algorithms Can Drift Toward Higher Prices
Modern pricing algorithms are built to maximise profit. They ingest live demand signals, remaining inventory, competitor fares, booking curves and, increasingly, shopper characteristics. They then adjust prices continuously.
Consider two airlines, X and Y, competing on the same route. Each runs its own independent algorithm. Early on, Airline X’s system tests a lower fare. Airline Y’s system detects the move and matches or undercuts it. Bookings rise briefly, but yields collapse. Both algorithms record the episode as unprofitable.
After several such cycles, Airline X’s algorithm updates its internal logic: aggressive undercutting triggers retaliation and destroys margin. It therefore becomes more cautious, preferring to hold or raise prices. Airline Y’s algorithm, observing the same pattern from its own side, reaches the identical conclusion. Over time the two systems settle into a stable, elevated price level.
No revenue manager at either airline has spoken to the other. No email or message has been exchanged. The outcome—higher fares that move in parallel—emerges purely from each algorithm’s independent attempt to maximise its own airline’s profit. Passengers pay more; the carriers may enjoy stronger short-term yields; yet the process has created the appearance, and potentially the substance, of coordination.
This is the core risk the industry now faces.
How Algorithmic Behaviour Becomes Unlawful Under Competition Law
Competition law in major jurisdictions (EU, UK, United States and many others) prohibits agreements, concerted practices or understandings that restrict competition. The classic requirement is a “meeting of minds”—some form of mutual commitment or coordination between independent companies.
Liability can arise in several distinct ways:
Intentional design or use
If an airline deliberately programmes an algorithm to coordinate with a competitor’s prices, or knowingly uses a tool for that purpose, the conduct is treated exactly as traditional cartel behaviour. The 2016 UK case against two online poster sellers confirmed that illegal offline conduct does not become legal simply because it is executed by software.
Shared algorithms or data hubs
When two or more airlines use the same third-party pricing platform, common data feed or joint algorithm in a way that allows alignment of prices or exchange of competitively sensitive information, authorities view this as a form of coordination. The sharing itself can supply the required meeting of minds.
Use of a competitor’s confidential information
Even without an explicit agreement, liability may attach if an airline’s algorithm can reasonably be expected to incorporate another carrier’s non-public data (for example through a shared vendor that fails to maintain proper firewalls). The anti-competitive outcome plus the foreseeable use of confidential information can be enough.
Inadequate oversight of autonomous systems
With agentic AI that makes pricing decisions with little or no human intervention, regulators are testing whether the absence of human communication still constitutes an unlawful understanding. Airlines that cannot explain how prices were set, or that operate on a “set it and forget it” basis, face heightened exposure. The EU AI Act already imposes risk-based obligations and requires meaningful human oversight for higher-risk systems.
Who can be held liable
The airline itself faces corporate fines that can reach 10 % of group turnover. In an industry where net margins frequently sit below 5 %, such penalties are severe. Individuals involved in designing, deploying or failing to supervise the systems may also face personal liability. Follow-on damages actions by passengers or competitors add further financial risk.
When the line is crossed
The critical moment is the combination of (a) parallel or elevated pricing outcomes that harm competition and (b) some element of coordination, shared infrastructure, foreseeable use of confidential data, or demonstrable failure to maintain independent control. Purely unilateral behaviour based solely on public information and an airline’s own data remains lawful; the difficulty lies in proving independence when outcomes look coordinated and oversight is weak.
How Airlines Can Mitigate the Risk
Effective mitigation is practical and achievable. Leading practice includes the following measures:
- Maintain a complete inventory of every pricing algorithm and third-party tool in use, documenting data inputs, decision rights and human override points.
- Prohibit any transmission of competitively sensitive information (current or planned prices, capacity intentions, detailed bid strategies) through algorithms or shared platforms.
- Avoid identical third-party pricing engines with direct competitors unless legal review confirms robust, auditable separation of data and models.
- Design systems with mandatory human oversight and the technical ability to intervene or override automated decisions.
- Require that every material pricing decision can be reconstructed and explained after the fact—both for internal governance and for regulators.
- Train commercial, revenue-management, IT and legal teams on the specific competition-law boundaries that apply to algorithmic pricing.
- Commission periodic independent audits that test for unexplained parallel price movements and verify that controls are operating.
- Insert strong contractual protections with technology vendors, including audit rights, data-separation warranties and allocation of regulatory liability.
- Keep a living compliance file that records what the airline knew, what controls it applied, and how it monitored outcomes.
These steps convert an opaque technical process into a managed, defensible one.
Additional Scenarios in Which Collusive-Like Outcomes Can Emerge
Beyond the basic learning loop, several other patterns are already attracting regulatory attention:
Parallel learning from public data alone
Both airlines scrape the same publicly available fares, schedules and load-factor indicators. Each algorithm independently concludes that matching rather than undercutting maximises long-term revenue. Prices rise in near-unison with no private communication.
Common third-party platform
Airlines X and Y both licence the same external dynamic-pricing service. Even if the vendor does not intentionally share confidential data, identical model structures and overlapping training signals can cause the two systems to converge on similar pricing logic.
Hub competition with connecting traffic
On overlapping local routes that feed each carrier’s hub, algorithms learn to protect high-yield connecting passengers by keeping local fares firm. The result is elevated point-to-point prices that move together.
Synchronised reaction to external shocks
A sharp rise in fuel prices or a sudden demand surge is processed by both systems. Each independently decides that an immediate price increase is optimal and that the competitor will do the same, producing rapid parallel movements.
Minimal-oversight agentic AI
Fully autonomous systems continuously re-optimise. Once they have internalised that discounting triggers retaliation, they rarely test lower prices. Human managers see strong yields and may delay intervention, allowing the high-price equilibrium to harden.
In each case the legal exposure grows if the airline cannot demonstrate that the behaviour was unilateral, based only on its own data and public information, and subject to effective human control.
Comparison with Traditional Revenue Management
Traditional revenue-management systems already produced parallel pricing. Analysts watched competitors’ published fares and adjusted their own buckets accordingly. The process was slower, more visible and easier to document.
Algorithmic systems differ in four material respects:
- Speed – adjustments occur continuously rather than in daily or hourly cycles.
- Granularity – prices can move in small increments across thousands of origin-destination and departure-date combinations.
- Opacity – the decision logic may reside inside complex models that are harder to interrogate.
- Autonomy – agentic systems can act with reduced human involvement.
The underlying competitive dynamic is not new; the velocity, scale and reduced transparency are. That is why regulators have intensified scrutiny.
Practical Compliance Posture for Airlines
A robust posture treats algorithmic pricing with the same seriousness as any other high-risk commercial process:
- Governance – clear ownership by a senior commercial or revenue-management executive, with legal and compliance representation.
- Documentation – written policies, data-flow maps, model inventories and decision logs retained for the relevant limitation periods.
- Technical controls – human override, access restrictions, and separation of any shared vendor environments.
- Monitoring – regular review of pricing outcomes for unexplained parallelism, supported by independent audits.
- Training and culture – repeated reinforcement that “the algorithm decided” is not an acceptable explanation.
- Vendor management – contractual rights to audit, require data firewalls and shift regulatory risk appropriately.
- Regulatory readiness – ability to respond rapidly to information requests with a coherent narrative of independence and control.
Airlines that embed these elements early reduce both the probability of infringement and the severity of any eventual penalty.
The Flipside: Improving Pricing and Profitability Legally
The same technological shift that creates risk also offers substantial lawful upside. Airlines can capture it by focusing on unilateral, efficiency-driven practices:
- Exploit proprietary data—own booking curves, website behaviour, ancillary conversion rates and customer lifetime value—to set prices that better reflect willingness to pay.
- Use NDC and Offers & Orders to construct personalised bundles (seat + bag + flexibility + extras) that create value through better matching rather than through aligned base fares.
- Tighten capacity and schedule decisions on the basis of the airline’s own network economics and cost structure.
- Run continuous, internal A/B tests of offer design, display logic and ancillary pricing; keep results confidential.
- Anchor prices to the carrier’s own costs, desired margins and differentiated product strengths; treat competitor fares as only one input among many.
- Invest in explainable models so that revenue managers retain genuine control and can satisfy both commercial and regulatory demands.
- Communicate clearly with customers when personalised pricing is used, reducing consumer-protection risk and building trust.
These approaches raise revenue by making the airline’s own decisions sharper and faster. They do not depend on predicting or matching a competitor’s next move—the zone that generates collusion exposure.
Summary and Conclusion
Pricing algorithms and agentic AI are powerful tools. They can independently learn that price wars destroy profit and therefore settle on higher fares. When that learning occurs in parallel across competing airlines, the outcome can look—and under certain conditions be treated as—unlawful coordination.
Liability turns on the presence of intentional design, shared infrastructure, foreseeable use of confidential information, or a failure to maintain independent human control. The penalties are severe relative to airline margins.
Yet the technology itself is not the enemy. Airlines that inventory their systems, impose rigorous oversight, document independence, and focus commercial effort on proprietary data and better offer design can improve yields legally and sustainably. Those that treat algorithms as black boxes, share tools carelessly, or operate without meaningful human accountability invite regulatory intervention.
The choice is straightforward. Use the new tools with eyes open, controls in place and a clear commercial strategy grounded in the airline’s own information. Done that way, the industry can realise the promised revenue gains while keeping both passengers and regulators onside. Done carelessly, the same tools become a source of collective loss—for carriers and for the travellers who ultimately pay the price.
