ICLE Comments on DOJ/FTC Guidance on Business Collaborations
I. Introduction
The International Center for Law & Economics (ICLE) submits these comments in response to the U.S. Department of Justice Antitrust Division (DOJ) and Federal Trade Commission (FTC) Joint Public Inquiry for Consideration of Guidance on Collaborations Among Competitors (Docket No. ATR-2026-0001).[1] Both agencies have acknowledged that new guidelines are needed. These comments therefore focus on what those guidelines should contain.
ICLE recommends that the agencies: (1) restore the general 20% combined-market-share safe harbor from the 2000 Antitrust Guidelines for Collaborations Among Competitors; (2) establish a 25% safe harbor for research-and-development (R&D) collaborations, along with a two-year grace period modeled on the European Union (EU) framework; (3) clarify that algorithmic-pricing liability requires evidence of coordination, not merely parallel use of software tools or platform policies; (4) establish safe harbors for procompetitive data-sharing arrangements, including artificial-intelligence (AI) training-data collaborations and cybersecurity-threat-intelligence sharing; and (5) confirm favorable rule-of-reason treatment for workforce credentialing, safety-standardization, and training collaborations that remain open and nondiscriminatory.
For 24 years, the 2000 Guidelines gave businesses a practical framework for evaluating competitor collaborations.[2] When the agencies withdrew those Guidelines in December 2024, they offered no replacement, transition period, or interim framework.[3] FTC Commissioner Melissa Holyoak warned in dissent that businesses would be left “grasping in the dark.”[4] Then-Commissioner Andrew Ferguson similarly argued that agency guidance must “properly inform[] the public of the Commission’s enforcement position” so businesses can plan with reasonable predictability.[5]
The resulting harm is not most visible in completed collaborations later challenged by enforcers. It appears in collaborations that never occur because firms narrow, delay, or abandon proposed arrangements in response to legal uncertainty. That effect is especially significant in research-intensive and globally competitive industries such as AI, semiconductors, and pharmaceuticals.
Recent developments in AI illustrate the problem. A March 2026 Lawfare analysis argued that antitrust uncertainty may discourage frontier AI firms from pursuing deeper joint safety initiatives, including collaborative model evaluations.[6] The article discussed a 2025 joint safety evaluation by OpenAI and Anthropic that was limited to publicly released models and suggested that collaboration involving unreleased systems may be constrained by antitrust concerns.[7] Anthropic has likewise stated that greater clarity on the antitrust treatment of AI-safety collaborations “would help determine whether and how AI labs can collaborate on safety standards.”[8]
Similarly, Bloomberg reported in April 2026 that OpenAI, Anthropic, and Google had begun sharing information through the Frontier Model Forum—the nonprofit they formed with Microsoft in 2023—to detect “adversarial distillation,” a technique used to replicate outputs from advanced AI systems.[9] According to the report, the companies remained uncertain about what information they could lawfully share under existing antitrust guidance and indicated that greater regulatory clarity would facilitate additional cooperation. The pattern is the same in both examples: firms seek to collaborate on objectives policymakers support, but legal uncertainty narrows the scope of permissible cooperation.
The agencies have themselves recognized the importance of clear guidance. Acting Assistant Attorney General Omeed Assefi stated in February 2026 that “vigorous and effective enforcement can only exist when the rules of the road are clearly outlined.”[10] Chairman Ferguson similarly observed that “businesses need transparency and predictability from enforcers more than ever.”[11] The case for new guidelines follows directly from those principles.
II. Restore and Expand Safe Harbors for Competitor Collaborations
Section 4.1 of the 2000 Guidelines explained the rationale for safe harbors directly, identifying “situations in which anticompetitive effects are so unlikely that the Agencies presume the arrangements to be lawful without inquiring into particular circumstances.” In 2024, by contrast, the FTC and DOJ argued that the 2000 Guidelines “risk creating safe harbors that have no basis in federal antitrust statutes.” That criticism misconstrued the nature of the safe harbors described in the Guidelines.
The 2000 Guidelines did not confer antitrust immunity or alter the substantive reach of the antitrust laws or Section 5 of the FTC Act. They recognized a basic economic reality: competitor collaborations are often procompetitive,[12] particularly when they involve firms with limited market power or facilitate productive joint activity such as R&D. The Guidelines identified categories of arrangements unlikely to harm competition and explained that the agencies were correspondingly unlikely to challenge them.[13] In doing so, they gave useful guidance to businesses seeking to comply with the law and to agency staff allocating limited enforcement resources.
Eliminating safe harbors does not improve enforcement precision. It risks directing agency resources toward collaborations unlikely to harm competition or consumers, while subjecting agreements among small competitors to the same uncertainty and scrutiny as arrangements involving firms with substantial market power. That approach can discourage efficient collaborations, particularly in research-intensive and globally competitive industries.
The agencies should therefore restore and expand clear safe-harbor guidance. As discussed below, that effort should include reinstating the 20% general safe harbor from the 2000 Guidelines, adopting a higher threshold for R&D collaborations that generate substantial knowledge spillovers, and creating a grace period for collaborations whose market shares later exceed the applicable threshold because the venture succeeded.
A. Restore the 20% General Safe Harbor
Section 4.2 of the 2000 Guidelines created a general safe harbor for collaborations in which the participants and the collaboration together accounted for no more than 20% of each relevant market in which competition could be affected. Even then, the agencies specified that the safe harbor did “not apply to agreements that are per se illegal, or that would be challenged without a detailed market analysis, or to competitor collaborations to which a merger analysis is applied.”[14]
The economic logic is straightforward. Firms with small combined market shares generally lack the market power needed to raise prices, restrict output, or exclude rivals in a durable way. A firm with a small market share that raises prices will typically lose sales to competitors rather than induce a market-wide price increase. For similar reasons, collaborations among firms with small aggregate market shares are unlikely to facilitate anticompetitive coordination or exclusion.
The Supreme Court’s decision in NCAA v. Alston reinforces the value of such safe harbors.[15] The Court confirmed that most agreements among competitors—including most joint-venture restrictions—are evaluated under the rule of reason, which requires a fact-intensive assessment of market conditions, competitive effects, and business justifications. Rule-of-reason analysis is expensive, time-consuming, and often unpredictable, even for arrangements unlikely to harm competition. Safe harbors reduce those costs by signaling that collaborations below a specified market-share threshold are unlikely to warrant enforcement scrutiny. In that sense, Alston makes safe harbors more valuable, not less.
The Biden administration justified the 2024 withdrawal of the Guidelines on the ground that the safety zones “have no basis in federal antitrust law.”[16] That rationale misunderstands the function of safe harbors. Safe harbors are enforcement-policy statements, not binding legal rules. The antitrust laws neither require nor prohibit them. The relevant policy question is whether the FTC and DOJ should devote enforcement resources to investigating collaborations among firms with low combined market shares. The answer is generally no. A 20% combined-share threshold remains a reasonable proxy for the point below which the structural conditions necessary for consumer harm are unlikely to exist.
ICLE therefore recommends restoring the general 20% safe harbor from Section 4.2 of the 2000 Guidelines. The same carveouts from the 2000 framework should also remain in place. The safe harbor should not apply to: (1) agreements that are per se illegal; (2) agreements the agencies would challenge without detailed market analysis; or (3) collaborations that function as mergers.
B. Adopt a Higher Safe-Harbor Threshold for R&D Collaborations
R&D collaborations warrant a higher safe-harbor threshold than ordinary competitor collaborations because they generate substantial knowledge spillovers. When one company invests in research, competitors and consumers often benefit indirectly through published findings, reverse engineering, employee mobility, or follow-on innovation. The investing firm therefore cannot capture the full social value of its research. As a result, private firms tend to invest less in R&D than would maximize overall welfare. Joint ventures can help correct that problem by allowing firms to share the costs, risks, and benefits of research projects that neither could efficiently undertake alone.
The economic literature supports this view. Morton Kamien, Eitan Muller, and Israel Zang show that R&D cooperation can increase total welfare relative to independent investment.[17] Empirical studies of European manufacturing firms likewise find that cooperation with other companies is associated with higher R&D spending.[18] In biopharmaceutical development, firms increasingly rely on alliances and joint ventures to share risk and combine complementary scientific and commercial capabilities, particularly when the average cost of bringing a new drug to market exceeds $1.3 billion per approved product.[19]
The EU recognized these dynamics in its revised R&D Block Exemption Regulation, which took effect July 1, 2023.[20] The regulation establishes a 25% combined-market-share safe harbor for joint R&D agreements among competitors. The European Commission explained that, below this threshold, the technical and economic benefits generated by R&D cooperation—including new products, improved technologies, and lower prices—will generally outweigh potential anticompetitive effects.[21] ICLE recommends that the FTC and DOJ adopt a comparable 25% threshold for R&D-specific safe harbors.
The international competitiveness implications are significant. American firms currently operate without a general safe-harbor regime, while EU firms benefit from clear thresholds under the 2023 Horizontal Block Exemption Regulations, including a 25% threshold for R&D agreements and a 20% threshold for specialization agreements.[22] A joint research venture between firms with a combined 22% market share may be lawful in both jurisdictions, but EU firms can confirm that quickly and predictably. American firms, by contrast, must either incur substantial legal-review costs or abandon the collaboration.
That disparity matters most in globally competitive industries such as AI, semiconductors, and pharmaceuticals, where firms make investment and collaboration decisions across jurisdictions. Regulatory clarity can influence where firms invest, conduct research, and form partnerships.
C. Adopt a Two-Year Grace Period
ICLE also recommends adopting a two-year grace period, consistent with the EU framework,[23] for collaborations whose participants exceed the applicable combined-market-share threshold after forming the arrangement.
Such a rule would prevent the safe harbor from creating perverse incentives against successful collaborations. Without a grace period, firms could face immediate legal uncertainty or enforcement risk simply because a productive collaboration succeeded in expanding output, improving products, or increasing market share. A grace period would instead allow firms time to adjust their arrangements or seek legal guidance while preserving the benefits of efficient cooperation.
III. Clarify Antitrust Treatment of Patent Pools and Joint Licensing
The agencies’ request for comment specifically asks whether joint licensing arrangements would benefit from additional guidance.[24] They would. Joint licensing among competitors—including patent pools, standard-essential-patent (SEP) platforms, and cross-licensing arrangements—is a paradigmatic example of a procompetitive horizontal collaboration of the type the 2000 Guidelines were designed to protect.
Properly structured joint-licensing arrangements can reduce transaction costs, mitigate royalty stacking, and accelerate the dissemination of standardized technologies, particularly in industries built around complex technical standards. The agencies should therefore reaffirm that patent pools limited to complementary patents and licensed on fair, reasonable, and nondiscriminatory (FRAND) terms generally warrant rule-of-reason treatment.
At the same time, the agencies should distinguish those arrangements from collective-bargaining groups among implementers. Existing antitrust doctrine already provides the tools necessary to distinguish procompetitive joint licensing from buyers’ cartels without categorical prohibitions or overbroad safe harbors.
A. Protect Procompetitive Patent Pools
Properly structured patent pools and licensing platforms can reduce transaction costs, mitigate royalty stacking, and accelerate the diffusion of standardized technologies.[25] Standard-setting organizations often develop technical standards that incorporate hundreds—or even thousands—of complementary patented technologies. Licensing those patents individually can require extensive negotiations, evaluation of large global patent portfolios, and specialized technical and legal counsel. Patent pools and licensing platforms offer a more efficient “one-stop-shop” alternative. Empirical research estimates that patent pools can save hundreds of millions of dollars in transaction costs relative to purely bilateral licensing.[26]
Competitive market forces constrain pool royalty rates when bilateral licensing remains available. In most patent pools, members retain the right to license their SEPs independently. An SEP is a patent that must be used to comply with a technical standard, such as a wireless-communications protocol. Because pool licenses and bilateral licenses operate as functional substitutes,[27] a pool operator cannot sustainably charge royalties above the value of available bilateral alternatives, adjusted for the efficiencies the pool creates. If a pool attempts to charge supra-competitive rates, rational licensees can negotiate directly with SEP holders, who remain bound by their FRAND licensing commitments.
The DOJ recognized these dynamics in its July 2020 Business Review Letter regarding the Avanci 5G Platform. DOJ concluded that the platform was “unlikely to harm competition”[28] and identified the preservation of independent bilateral licensing as a key safeguard against anticompetitive effects.[29] In a November 2022 letter signed by 25 legal academics, economists, former federal judges, and other antitrust and intellectual-property experts, ICLE and others urged the agencies to preserve that approach.[30]
New guidelines should confirm that properly structured patent pools are presumptively procompetitive and should be analyzed under the rule of reason. That treatment should apply to pools limited to essential and complementary patents, licensed on FRAND terms, and structured to preserve participants’ ability to license bilaterally outside the pool.
B. LNGs Are Not Patent Pools
A separate question is whether collective bargaining by implementers—sometimes called licensing-negotiation groups (LNGs)—should receive safe-harbor treatment analogous to patent pools. The European Commission initially proposed such a safe harbor in its draft revised Technology Transfer Guidelines before ultimately retreating from that approach[31] The FTC and DOJ should likewise decline to adopt one.
Proponents of LNG safe harbors often argue that LNGs are simply the mirror image of patent pools. That analogy is flawed. Patent pools involve the collective licensing of complementary patents—separately owned technologies that must be combined to implement a standard. By aggregating those rights, pools can reduce transaction costs and mitigate royalty stacking,[32] which occurs when multiple overlapping royalty demands increase the total cost of implementing a technology standard.
LNGs operate differently. Rather than coordinating the sale of complementary technologies, they coordinate the purchase of licenses among implementers that would otherwise negotiate independently. In practice, LNGs involve buyers collectively setting the price they are willing to pay for technology licenses.[33] That structure more closely resembles a buyers’ cartel than a patent pool. Treating LNGs as equivalent to patent pools collapses the important distinction between coordinating complements and coordinating substitutes, and improperly extends the efficiency rationale for pools to fundamentally different conduct.
Buyers’ cartels can be as harmful as sellers’ cartels. Just as a sellers’ cartel can inflate prices above competitive levels, a buyers’ cartel can suppress prices below competitive levels. In the context of intellectual property, suppressed royalties can reduce returns to innovation and weaken incentives to invest in future technologies.[34] U.S. antitrust authorities have long recognized this principle. In the music-licensing context, for example, the DOJ filed a statement of interest explaining that collective rate-negotiation strategies could constitute per se unlawful buyers’ cartels.[35] More recently, a senior DOJ official criticized the European Commission’s support for an automotive LNG as “unfortunate” and difficult to reconcile with sound competition-law principles, warning that similar arrangements would likely be treated as per se unlawful buyers’ cartels under U.S. law.[36]
ICLE therefore recommends that any revised guidelines make three points clear. First, properly structured patent pools and SEP platforms—including those limited to essential and complementary patents, licensed on FRAND terms, and preserving participants’ rights to license bilaterally—are presumptively procompetitive and should be analyzed under the rule of reason, consistent with the 2020 Avanci Business Review Letter. Second, the agencies should not extend safe-harbor treatment to implementer collective-bargaining arrangements that lack the structural characteristics that make patent pools procompetitive. Third, existing antitrust principles, applied case by case, are sufficient to address potential harms from joint licensing without categorical prohibitions or overbroad safe harbors.
IV. Distinguish Algorithmic Coordination from Lawful Conduct
The agencies’ request for comment specifically asks whether new technologies and business models—including algorithmic pricing and data sharing—would benefit from additional guidance[37] They would. Clearer guidelines are needed to distinguish lawful parallel conduct from unlawful coordination.
The key question should be whether competitors share nonpublic, competitively sensitive information or otherwise coordinate their conduct. Liability should not turn merely on the use of algorithms, common software tools, or uniform platform policies. Algorithms often automate lawful conduct, such as responding to publicly available prices and changing market conditions, and can generate substantial consumer benefits.
New guidelines should therefore focus on identifying concrete mechanisms of coordination, distinguishing public from nonpublic information, and preserving the distinction between vertical platform policies and horizontal agreements among competitors.
A. A Framework for Evaluating Algorithmic Pricing
Algorithmic pricing can arise in at least three distinct situations that require different antitrust treatment. First, competitors may agree to use a shared pricing algorithm to coordinate prices. In that scenario, the algorithm itself is legally irrelevant; the unlawful agreement is the antitrust violation, and existing law already addresses it.
Second, competing firms may each share competitively sensitive information with a common third-party platform that generates pricing recommendations for all participants. In that circumstance, the platform can function as a coordinating mechanism because each participant receives recommendations informed by rivals’ nonpublic data, even absent direct communication among competitors.
Third, independent algorithms may arrive at similar pricing outcomes simply because they respond to the same market conditions and economic signals. That form of conscious parallelism is not unlawful under U.S. antitrust law absent an agreement among competitors.
The proposed consent decree in the DOJ’s RealPage case offers a useful framework for distinguishing lawful from unlawful conduct.[38] The decree targets the use of competitors’ nonpublic, competitively sensitive information in generating pricing recommendations, while preserving firms’ ability to rely on their own internal data and publicly available market information. As the Competitive Impact Statement explains, the decree permits pricing recommendations “solely based on a property owner’s own nonpublic data or data available to the general public,” reflecting “the principle that firms should not make pricing decisions using insight drawn from their competitors’ nonpublic, competitively sensitive data.”[39]
New guidelines should adopt that distinction as their organizing principle.
B. Separate Public and Sensitive Data
An algorithm that monitors publicly posted competitor prices and adjusts accordingly merely automates conduct that human pricing analysts have long performed lawfully. As former FTC Acting Chairman Maureen Ohlhausen observed, automated business practices should generally be evaluated by asking whether the same conduct would be lawful if performed manually—her “guy named Bob” test.[40] The antitrust concern arises when competitors share nonpublic, competitively sensitive information through a common platform. New guidelines should make that distinction explicit.
Precision matters. In the RealPage litigation, the problematic information was not publicly visible asking rents posted on apartment-listing websites. It instead involved nonpublic transactional data, including executed lease prices, concessions and discounts, renewal terms, and forward-looking occupancy projections.[41] The distinction between public pricing information and confidential transactional data is critical. Not all commercially relevant information is sufficiently sensitive to create antitrust concerns.
The hotel-pricing software at issue in Gibson v. Cendyn illustrates the other side of the line.[42] There, each hotel’s pricing recommendations were generated using the hotel’s own internal data combined with publicly available competitor rates collected from travel websites. The software did not pool or share one hotel’s proprietary data with competing hotels. The 9th U.S. Circuit Court of Appeals therefore affirmed dismissal of the antitrust claims, holding that independent use of the same pricing algorithm does not itself constitute a restraint of trade, even when competitors know they are using the same tool.
That outcome follows naturally from the distinction between public and nonpublic information. When a pricing system excludes competitors’ confidential data, the factual predicate for hub-and-spoke liability is absent.
C. Require Proof of Coordination
As former Deputy Assistant Attorney General Roger Alford explained, “in the absence of evidence of concerted action, we cannot presume the simple use of pricing algorithms is an antitrust violation.”[43] That principle remains correct. The use of common software tools, standing alone, does not establish unlawful coordination.
The Gibson v. Cendyn litigation illustrates the point. As discussed above, the software architecture in that case excluded competitors’ confidential data.[44] Each hotel’s pricing recommendations were generated from its own internal information and publicly available market data, rather than from competitors’ nonpublic information. When a platform operates this way, the factual predicate for hub-and-spoke liability is absent because the platform is not functioning as a conduit for sharing competitively sensitive information among rivals.
New guidelines should identify the plus factors necessary to transform parallel adoption of a common pricing tool into a hub-and-spoke conspiracy.[45] Relevant evidence could include: (1) coordinated adoption decisions among competitors; (2) agreements not to deviate from algorithmic recommendations; (3) exchange of competitors’ nonpublic information through the platform to generate pricing recommendations; and (4) software features specifically designed to align prices or penalize deviation from recommended pricing.
The agencies should not challenge algorithmic-pricing conduct without first identifying a concrete mechanism of coordination among competitors.
D. Preserve the Vertical-Horizontal Distinction
Another area requiring clearer guidance is the treatment of platform pricing policies, including most-favored-nation (MFN) clauses, price-parity provisions, and “fair pricing” policies. Recent litigation increasingly characterizes these arrangements as hub-and-spoke conspiracies facilitating horizontal coordination among sellers.[46] That framing risks collapsing ordinary vertical contracting into per se unlawful horizontal price-fixing.
A platform that adopts uniform pricing terms across its seller agreements engages in a common form of vertical contracting. Sellers subject to these policies do not gain access to competitors’ costs, pricing strategies, or proprietary information. Nor does widespread use of a common contractual term transform a series of bilateral vertical agreements into a horizontal conspiracy. A franchise system’s uniform quality standards, for example, do not create a conspiracy among franchisees simply because all franchisees operate under similar contractual obligations.
MFN provisions are longstanding commercial tools with recognized procompetitive justifications.[47] Judge Richard Posner described MFNs as “standard devices by which buyers try to bargain for lower prices” and “the sort of conduct that the antitrust laws seek to encourage.”[48] The economic literature likewise finds ambiguous competitive effects. The strongest empirical studies report either no statistically significant medium-run price effects from eliminating MFNs or effects that are small and temporary.[49]
United States v. Apple illustrates the substantial evidentiary showing required to establish hub-and-spoke liability.[50] The case involved extensive evidence of horizontal coordination among publishers, direct communications regarding collective strategy, and active orchestration by Apple itself. More broadly, Apple demonstrates that hub-and-spoke liability requires proof of a horizontal agreement among the competing “spokes”—the “rim” of the wheel.[51] If liability was difficult to establish even where a platform actively coordinated supplier conduct, it should not arise absent evidence of horizontal agreement among a platform’s participants.[52] Uniform contractual terms alone—even when imposed by a dominant platform—cannot supply the missing rim.
New guidelines should therefore reaffirm three principles. First, hub-and-spoke liability requires proof of a horizontal agreement connecting competing firms.[53] Second, conscious parallelism, even when facilitated by common tools or market conditions, is not itself a conspiracy.[54] Third, vertical contract terms between a platform and its participants are analytically distinct from horizontal agreements among those participants.
E. Avoid Guidance That Chills Procompetitive Pricing Technologies
Guidelines that treat common algorithmic tools or platform pricing policies as presumptive hub-and-spoke coordination would jeopardize substantial consumer benefits. Dynamic pricing can improve allocative efficiency by matching supply more closely to demand. Empirical research finds that dynamic airline pricing increases output and overall welfare, while rideshare surge pricing increases rider surplus and disproportionately benefits lower-income riders.[55] Third-party pricing tools can also reduce barriers to entry by giving smaller firms access to sophisticated pricing capabilities previously available only to larger competitors. The Organisation for Economic Co-operation and Development (OECD) has recognized that pricing algorithms can help smaller firms gain market insights and compete more effectively.[56]
The Mercatus Center’s 2025 review likewise identified several consumer benefits from algorithmic pricing, including reduced shortages, improved product availability for higher-valuation consumers, and greater responsiveness to changing market conditions.[57] The review concluded that overbroad restrictions on algorithmic pricing would eliminate those benefits without addressing the narrower coordination risks the evidence supports.
Platform pricing policies can likewise serve legitimate procompetitive purposes. The Supreme Court has long recognized that vertical restraints may prevent free riding and protect investment in valuable services and infrastructure.[58] Online platforms often invest heavily in search functionality, product reviews, fraud prevention, payment processing, and logistics. When platforms cannot prevent sellers from using those services while undercutting prices elsewhere, they may shift toward less transparent alternatives such as algorithmic steering, advertising-based rankings, or delisting practices that may prove more harmful to consumers and more difficult to administer.
The empirical literature further underscores the need for case-specific analysis. The competitive effects of algorithmic pricing vary across markets and business models. The same technology that may soften competition in some settings can intensify it in others, as the German gasoline-market study by Stephanie Assad and coauthors demonstrates.[59] New guidelines should therefore focus on concrete evidence of coordination rather than broadly condemning algorithmic pricing technologies or platform pricing policies.
V. Provide Clear Safe Harbors for Procompetitive Data Sharing
Data sharing presents some of the most consequential unresolved questions in modern antitrust law. The agencies’ request for comment specifically identifies AI training data, cybersecurity-threat intelligence, and benchmarking arrangements as priority areas. Each raises distinct competitive considerations and requires a different analytical framework.
The key distinction is not whether firms share data, but what kind of data they share and whether the arrangement facilitates coordination among competitors. Existing antitrust doctrine already distinguishes between exchanges of current competitively sensitive information, which may facilitate collusion, and exchanges of aggregated, historical, anonymized, or technical information, which often generate substantial procompetitive benefits.
New guidelines should build on those distinctions by adopting targeted safe harbors and clear governance safeguards to distinguish legitimate information sharing from unlawful coordination.
A. Governance Safeguards for AI Training Data
Building competitive AI systems requires large amounts of training data. Research suggests that AI development is often hindered by fragmented or siloed datasets, which increase the time and labor needed to clean, label, and organize training inputs. Standardized or pooled data architectures, by contrast, can improve data availability, governance, and model-development efficiency.[60] A 2025 Mercatus Center working paper likewise found that open-source AI models and commercial data-licensing markets provide smaller firms meaningful access routes to training data, reducing the likelihood that data pooling itself will foreclose competition.[61]
These findings support a data-sharing safe harbor focused on governance safeguards rather than categorical restrictions on pooling. The European Commission’s 2023 Horizontal Guidelines adopt a similar distinction. They treat exchanges of current competitively sensitive information as presumptively problematic because they can facilitate coordination of future conduct, while treating historical or aggregated data managed by an independent intermediary as generally less concerning.[62]
The relevant competition concern is not that firms gain access to larger training datasets. Competition in AI markets generally centers on model architectures, computing resources, training methods, and post-training capabilities—not merely on access to raw inputs. The concern instead is that firms could use a purported “training-data pool” to exchange current pricing, output, capacity, or forward-looking business information that would otherwise be unlawful to share directly. Likewise, a jointly trained and jointly deployed model could become a coordination mechanism during real-world use.
Broad restrictions on data pooling would likely do more harm than good. Effective AI training often requires extremely large datasets, and startups may need collaborative arrangements to achieve the scale necessary to compete with incumbents.
The agencies should therefore adopt a safe harbor for data-sharing arrangements that satisfy four conditions: (1) an independent third party administers the arrangement; (2) shared data is aggregated or anonymized; (3) current pricing, output, or forward-looking business-planning information is excluded; and (4) access is available to market participants on nondiscriminatory terms.
B. Protect Cybersecurity Information Sharing
In April 2014, the FTC and DOJ issued a joint policy statement explaining that cybersecurity information—such as technical data about hacking methods, attack signatures, and defensive countermeasures—is fundamentally different from competitively sensitive commercial information and therefore “is not likely to raise antitrust concerns.”[63] Congress later reinforced that principle through the Cybersecurity Information Sharing Act of 2015, which established a statutory antitrust exemption for private entities sharing cyber-threat indicators or providing cybersecurity assistance under the act.[64]
The core distinction recognized in the 2014 policy statement remains sound and should be reaffirmed in new guidelines. Technical threat intelligence is categorically different from pricing, output, or other commercial data that could facilitate coordination among competitors. A firm that learns about a new malware signature or network vulnerability through an industry-sharing group gains information relevant to cybersecurity defense, not information that would help coordinate prices or allocate markets.
New guidelines should therefore confirm that cybersecurity-threat-intelligence sharing—limited to technical indicators and expressly excluding current commercial information—does not raise antitrust concerns, regardless of participants’ market shares.
C. Use Existing Information-Sharing Frameworks as the Model
The governance safeguards proposed above for AI training data and cybersecurity intelligence are not novel. They build on information-sharing frameworks the agencies have previously endorsed. Before withdrawing the health-care antitrust policy statements in 2023, the FTC and DOJ recognized a safe harbor for certain health-care benchmarking and information-sharing arrangements. To qualify, the data had to be collected by an independent third party, be more than three months old, include at least five contributors, ensure that no participant supplied more than 25% of the inputs, and be reported only in aggregated form.[65]
Those conditions reflected a broader antitrust principle: aggregated and historical data exchanges are generally less likely to facilitate coordination than direct exchanges of current competitively sensitive information. New guidelines should codify those principles as the baseline framework for lawful information sharing, with the AI and cybersecurity safe harbors discussed above serving as targeted applications.
For industry benchmarking more broadly, the EU framework for standardization agreements provides a useful model.[66] The key safeguards are transparency, open participation on equal terms, voluntary participation, limits on the exchange of competitively sensitive information, and broad access to the resulting benchmarks. These conditions help ensure that benchmarking arrangements serve legitimate informational purposes rather than becoming vehicles for coordination among competitors.
VI. Support Procompetitive Workforce Standards
The agencies’ January 2025 Antitrust Guidelines for Business Activities Affecting Workers correctly recognize that collaborations involving professional credentials, safety certifications, and workforce-training standards are generally procompetitive.[67] New guidelines should reaffirm and extend that principle beyond the specific arrangements addressed in the worker guidelines.
The economic rationale is straightforward. Professional credentialing and standardized training reduce information asymmetries between workers and employers by providing independent, verifiable signals of competence. That, in turn, can improve matching efficiency and increase labor mobility by making credentials portable across employers and regions.
Courts have likewise recognized that workforce-related restraints embedded in broader collaborations may warrant rule-of-reason treatment. In Aya Healthcare Services, Inc. v. AMN Healthcare, Inc., for example, the 9th U.S. Circuit Court of Appeals analyzed a non-solicitation provision as an ancillary restraint reasonably related to a broader procompetitive staffing arrangement, rather than as a naked restraint on competition.[68]
The limiting principle is clear. Collaborative standard-setting and credentialing arrangements should be treated as presumptively procompetitive when they are reasonably related to legitimate quality or safety objectives, open to qualified participants on equal terms, and not used to fix wages, allocate employees, or exclude rivals through unjustified restrictions. New guidelines should confirm rule-of-reason treatment for arrangements that satisfy those conditions.
VII. Clarify Rule-of-Reason Analysis for Joint Ventures
In NCAA v. Alston, the Supreme Court reaffirmed that most agreements among competitors—including most joint-venture restraints—are analyzed under the rule of reason, which requires a fact-specific assessment of market conditions, competitive effects, and asserted business justifications.[69] Rule-of-reason analysis is often expensive, time-consuming, and unpredictable, even for collaborations unlikely to harm competition. Safe harbors reduce those costs by signaling that low-share collaborations are unlikely to raise competitive concerns. As discussed above, Alston makes such guidance more valuable, not less.
The 1st U.S. Circuit Court of Appeals’ decision in United States v. American Airlines Group Inc. further illustrates the importance of focusing on actual competitive effects rather than labels.[70] The court upheld the government’s challenge to the Northeast Alliance joint venture between American Airlines and JetBlue after finding that the arrangement reduced competition on overlapping routes and decreased output in the form of fewer flights available to consumers. The airlines’ asserted efficiencies failed because they lacked reliable evidentiary support and could have been achieved through less restrictive means.
The key lesson for agency guidance is that calling an arrangement a “joint venture” does not entitle it to favorable treatment. The relevant inquiry is whether the collaboration harms competition in practice. Conversely, collaborations among firms with small combined market shares generally lack the ability to reduce output, raise prices, or exclude rivals in a durable way. Any procompetitive presumption for low-share collaborations should therefore rest on market structure and competitive realities, not formal labels.
The American Airlines litigation also highlighted continuing uncertainty surrounding evidentiary burdens in rule-of-reason cases. American Airlines petitioned the Supreme Court for review, arguing that the circuits apply inconsistent standards at the first step of rule-of-reason analysis.[71] Although the Court denied certiorari,[72] the case underscores the need for clearer agency guidance on what evidence of anticompetitive effects is necessary to satisfy the plaintiff’s initial burden in joint-venture cases.
VIII. Close the U.S.-EU Guidance Gap
American businesses evaluating competitor collaborations currently lack any general safe-harbor regime. By contrast, the European Union’s 2023 Horizontal Block Exemption Regulations provide market-share safe harbors for certain research-and-development (R&D) agreements (25% combined market share) and specialization agreements (20% combined market share).[73] The European Commission’s 2023 Horizontal Guidelines also establish a soft safe harbor for certain sustainability-standardization agreements that satisfy specified conditions.[74] In both the R&D and specialization contexts, the EU further provides a two-year grace period when market shares later exceed the applicable threshold.[75] The United States offers no comparable framework.
That disparity has direct competitive consequences. In industries where collaboration decisions are made globally—including artificial intelligence (AI), semiconductors, pharmaceuticals, and advanced manufacturing—the ex ante legal costs of evaluating a collaboration in the United States are now materially higher than in the EU. This difference does not reflect a substantive divergence in the legality of many arrangements. A joint R&D venture between firms with a combined 22% market share may be lawful in both jurisdictions. The difference is that EU firms can determine that more quickly and predictably, while American firms must either incur substantial legal-review costs or forgo the collaboration.
The EU framework therefore makes the costs of the current U.S. guidance gap concrete. American firms face systematically higher collaboration costs than their European counterparts in precisely the industries where international competition and innovation matter most.
IX. Recommendations
ICLE recommends that the agencies incorporate the following principles into new guidelines.
General Safe Harbor. The agencies should not challenge a competitor collaboration when the collaboration and its participants collectively account for no more than 20% of any relevant affected market. The carveouts from the 2000 Guidelines should remain in place. The safe harbor should not apply to: (1) per se unlawful conduct; (2) agreements the agencies would challenge without detailed market analysis; or (3) collaborations that function as mergers.
R&D Safe Harbor. The agencies should not challenge an R&D collaboration when participants’ combined market share does not exceed 25% in any relevant affected market. This threshold reflects the distinctive efficiencies associated with R&D cooperation, including the pooling of complementary assets, expertise, and investment.
Grace Period. A collaboration that exceeds the applicable market-share threshold because participants’ shares increased after the arrangement was formed should retain safe-harbor protection for two calendar years after the threshold is exceeded.
Algorithmic Pricing. Algorithmic-pricing practices should be analyzed under the rule of reason. Independent adoption of pricing software, absent an agreement, shared platform, or exchange of current competitively sensitive information among competitors, does not constitute a Sherman Act violation. The agencies should not challenge algorithmic-pricing conduct without identifying a concrete mechanism of coordination.
Data-Sharing Safe Harbor. Data-sharing arrangements should be presumptively lawful regardless of market share when four conditions are satisfied: (1) an independent third party administers the arrangement; (2) only aggregated or anonymized data is shared; (3) current pricing, output, capacity, or business-planning information is excluded; and (4) access is available to market participants on nondiscriminatory terms.
Cybersecurity Threat Intelligence. Sharing technical cyber-threat indicators, attack signatures, and defensive countermeasures should not raise antitrust concerns, regardless of market share, provided the sharing excludes current commercial pricing, output, or other competitively sensitive business information.
Labor-Market Standard-Setting. Collaborative development of professional credentials, training standards, and workplace-safety standards should receive favorable rule-of-reason treatment when: (1) the standards are reasonably related to legitimate quality or safety objectives; (2) participation is open to qualified parties on nondiscriminatory terms; and (3) the arrangements do not fix wages, allocate employees, or restrict labor supply through exclusionary criteria.
[1] Fed. Trade Comm’n, Press Release, Federal Trade Commission and Department of Justice Seek Public Comment for Guidance on Business Collaborations (Feb. 23, 2026), https://www.ftc.gov/news-events/news/press-releases/2026/02/federal-trade-commission-department-justice-seek-public-comment-guidance-business-collaborations [hereinafter Request for Comment].
[2] U.S. Dep’t of Justice & Fed. Trade Comm’n, Antitrust Guidelines for Collaborations Among Competitors (2000), https://www.ftc.gov/sites/default/files/documents/public_events/joint-venture-hearings-antitrust-guidelines-collaboration-among-competitors/ftcdojguidelines-2.pdf [hereinafter 2000 Guidelines].
[3] U.S. Dep’t of Justice & Fed. Trade Comm’n, Announcement, Justice Department and Federal Trade Commission Withdraw Guidelines for Collaboration Among Competitors (Dec. 11, 2024), https://www.justice.gov/atr/media/1380001/dl?inline.
[4] Fed. Trade Comm’n, Statement of Comm’r Melissa Holyoak on the Withdrawal of the Antitrust Guidelines for Collaborations Among Competitors (Dec. 11, 2024), https://www.ftc.gov/system/files/ftc_gov/pdf/holyoak-collaboration-guidelines-withdrawal-statement.pdf.
[5] Fed. Trade Comm’n, Statement of Comm’r Andrew Ferguson on the Withdrawal of the Antitrust Guidelines for Collaborations Among Competitors (Dec. 11, 2024), https://www.ftc.gov/system/files/ftc_gov/pdf/collaborations-guidance-withdrawal-ferguson_-statement.pdf.
[6] Nicholas Felstead, How Antitrust Can Promote AI Safety Collaborations, Lawfare (Mar. 5, 2026), https://www.lawfaremedia.org/article/how-antitrust-can-promote-ai-safety-collaborations.
[7] OpenAI, Findings from a Pilot Anthropic-OpenAI Alignment Evaluation Exercise (Aug. 26, 2025), https://openai.com/index/openai-anthropic-safety-evaluation.
[8] Anthropic, Comment Letter on AI Accountability Policy (June 6, 2023), https://www-cdn.anthropic.com/257e6352c677beeffcbce24233211887173a41dc/2023.06.06-Anthropic_NTIA_Comment_v2.pdf.
[9] Shirin Ghaffary & Maggie Eastland, OpenAI, Anthropic, Google Unite to Combat Model Copying in China, Bloomberg (Apr. 6, 2026), https://www.bloomberg.com/news/articles/2026-04-06/openai-anthropic-google-unite-to-combat-model-copying-in-china (“The three companies are uncertain about the scope of information they can legally share under existing antitrust guidelines” and have “indicated they would benefit from greater regulatory clarity from the US government.”).
[10] Request for Comment, supra note 1.
[11] Id.
[12] 2000 Guidelines, supra note 2, at 25.
[13] In that regard, the 2000 Guidelines resembled other agency guidance documents, such as the 1997 Horizontal Merger Guidelines, cited in the 2000 Guidelines. Id. at 26 n.3. The 2010 Horizontal Merger Guidelines likewise identified two categories of mergers “unlikely to have adverse competitive effects and ordinarily require no further analysis,” even though they did not use the term “safety zone.” U.S. Dep’t of Justice & Fed. Trade Comm’n, Horizontal Merger Guidelines § 5.3 (2010), https://www.justice.gov/atr/file/810276/dl?inline.
[14] 2000 Guidelines, supra note 2, at 26–27 (internal citations omitted).
[15] NCAA v. Alston, 594 U.S. 69 (2021).
[16] Fed. Trade Comm’n, Statement of Comm’r Alvaro M. Bedoya Regarding the Withdrawal of the Antitrust Guidelines for Collaborations Among Competitors (Dec. 11, 2024), https://www.ftc.gov/system/files/ftc_gov/pdf/bedoya-statement-regarding-withdrawal-collaboration-guidelines.pdf.
[17] Morton I. Kamien, Eitan Muller & Israel Zang, Research Joint Ventures and R&D Cartels, 82 Am. Econ. Rev. 1293 (1992).
[18] Bojan Lalic, Tanja Todorovic, Nenad Medic, Branislav Bogojevic, Danijela Ciric & Ugljesa Marjanovic, The Impact of Inter-Organizational Cooperation on R&D Expenditure of Manufacturing Companies, 39 Procedia Manufacturing 1401 (2019).
[19] See, e.g., Neil Lesser & Matt Hefner, R&D Partnerships—Partnering for Progress: How Collaborations Are Fueling Biomedical Advances, Drug Dev. & Delivery (Nov.–Dec. 2017), https://drug-dev.com/rd-partnerships-partnering-for-progress-how-collaborations-are-fueling-biomedical-advances; see also Andrew Mulcahy, Stephanie Rennane, Daniel Schwam, Reid Dickerson, Lawrence Baker & Kanaka Shetty, Use of Clinical Trial Characteristics to Estimate Costs of New Drug Development, 8 JAMA Network Open e2453275 (Jan. 6, 2025), https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2828689 (reporting a mean drug-development cost of $1.3 billion and a median cost of $708 million).
[20] Commission Regulation (EU) 2023/1066 of 1 June 2023 on the Application of Article 101(3) of the Treaty on the Functioning of the European Union to Certain Categories of Research and Development Agreements, 2023 O.J. (L 143) 9.
[21] Id. rec. 5 (“Below a certain level of market power, it can in general be presumed, for the application of Article 101(3) of the Treaty, that the positive effects of research and development agreements will outweigh any negative effects on competition.”).
[22] Commission Regulation 2023/1066, supra note 20; Commission Regulation 2023/1067, 2023 O.J. (L 143) 20 (EU) (specialization agreements).
[23] Both the R&D and specialization block exemptions include two-year grace periods after market shares exceed the relevant threshold.
[24] Request for Comment, supra note 1. (“What topics would benefit from additional guidance—for example, joint licensing arrangements?”).
[25] Written Submissions of the International Center for Law & Economics as Intervener at ¶ 2.1.1, Tesla, Inc. v. InterDigital Patent Holdings, Inc. & Avanci, LLC, UKSC/2025/0058 (U.K. Sup. Ct. Mar. 16, 2026) [hereinafter ICLE Tesla v. Avanci Submissions].
[26] Robert P. Merges & Michael Mattioli, Measuring the Costs and Benefits of Patent Pools, 78 Ohio St. L.J. 281 (2017) (presenting the first empirically based estimate of patent-pool transaction-cost savings and finding that pools can save hundreds of millions of dollars relative to bilateral-licensing counterfactuals).
[27] ICLE Tesla v. Avanci Submissions, supra note 25, ¶ 6.3.
[28] U.S. Dep’t of Justice, Business Review Letter Re: Avanci 5G Platform 2–3 (July 28, 2020) [hereinafter DOJ Avanci Business Review Letter] (“After soliciting input from a range of stakeholders in the automotive and telecommunications industries, including potential licensors and licensees, conducting an independent review, and considering our prior guidance and reviews of other patent pools, we conclude that, on balance, Avanci’s proposed 5G Platform is unlikely to harm competition.”).
[29] Id. at 21 (“There is no single correct way to calculate a reasonable royalty in the FRAND context.”).
[30] Letter from Adam Mossoff, Geoffrey A. Manne, Gus Hurwitz et al. to Jonathan Kanter, Assistant Att’y Gen., Re: Support for the Avanci Business Review Letter 1 (Nov. 30, 2022) [hereinafter Mossoff/ICLE Kanter Letter]. The letter was signed by 25 experts, including former Federal Circuit Judges Paul R. Michel and Kathleen M. O’Malley and former D.C. Circuit Judge Douglas H. Ginsburg, and characterized the 2020 Avanci business review letter as “a legally sound and evidence-based approach in applying antitrust law to innovative commercial institutions like the Avanci patent pool that facilitate the efficient commercialization of new standardized technologies in the fast-growing mobile telecommunications sector to the benefit of innovators, implementers, and consumers alike.”
[31] Communication from the Commission—Approval of the Content of a Draft for a Commission Regulation on the Application of Article 101(3) of the Treaty on the Functioning of the European Union to Categories of Technology Transfer Agreements and a Draft for Commission Guidelines on the Application of Article 101 of the Treaty to Technology Transfer Agreements, 2025 O.J. (C/2025/5024), § 4.5 (Sept. 16, 2025); see also Comments of the International Center for Law & Economics on the Draft Revised Technology Transfer Block Exemption Regulation and Technology Transfer Guidelines 4–7 (Oct. 23, 2025) [hereinafter ICLE TTBER Comments].
[32] See Josh Lerner & Jean Tirole, Efficient Patent Pools, 94 Am. Econ. Rev. 691 (2004); Carl Shapiro, Navigating the Patent Thicket: Cross Licenses, Patent Pools, and Standard Setting, 1 Innovation Pol’y & Econ. 119, 134 (2000) (“In many respects, a patent pool (much like a package license) is the purest solution to the complements problem described above and analyzed in the appendix.”).
[33] See, e.g., Igor Nikolic, Licensing Negotiation Groups for SEPs: Collusive Technology Buyers Arrangements? Their Pitfalls and Reasonable Alternatives, Les Nouvelles 226 (2021).
[34] Id.
[35] Statement of Interest of the United States, Global Music Rights, LLC v. Radio Music License Comm., Inc., No. 2:16-cv-09051-TJH-AS (C.D. Cal. Dec. 5, 2019), ECF No. 111.
[36] Khushita Vasant, EU Guidance on Carmakers’ SEP Licensing “Unfortunate,” US DOJ’s Kallay Says, MLex (Oct. 10, 2025), https://www.mlex.com/mlex/amp/articles/2398760.
[37] Request for Comment, supra note 1.
[38] U.S. Dep’t of Justice, Press Release, Justice Department Requires RealPage to End the Sharing of Competitively Sensitive Information and Alignment of Pricing Among Competitors (Nov. 24, 2025), https://www.justice.gov/opa/pr/justice-department-requires-realpage-end-sharing-competitively-sensitive-information-and.
[39] Competitive Impact Statement at 10–11, United States v. RealPage, Inc., No. 1:24-cv-00710 (M.D.N.C. Nov. 24, 2025), ECF No. 160, https://www.justice.gov/atr/media/1419471/dl.
[40] Maureen K. Ohlhausen, Acting Chairman, Fed. Trade Comm’n, Should We Fear the Things That Go Beep in the Night? Some Initial Thoughts on the Intersection of Antitrust Laws and Algorithmic Pricing 10 (May 23, 2017), https://www.ftc.gov/system/files/documents/public_statements/1220893/ohlhausen_-_concurrences_5-23-17.pdf (“Is it ok for a guy named Bob to collect confidential price strategy information from all the participants in a market, and then tell everybody how they should price? If it isn’t ok for a guy named Bob to do it, then it probably isn’t ok for an algorithm to do it either.”).
[41] Competitive Impact Statement, supra note 39, at 7–8 (identifying the problematic data as “rental applications, executed new leases, renewal offers and acceptances, and occupancy estimates and projections”).
[42] Brief of the International Center for Law & Economics as Amicus Curiae at 6, Gibson v. Cendyn Grp., LLC, No. 23-16463 (9th Cir. Dec. 26, 2024), https://laweconcenter.org/resources/icle-brief-to-the-9th-circuit-in-gibson-v-cendyn (“[T]here is no allegation here that Rainmaker’s pricing recommendations to one subscriber are based on the confidential information of another subscriber.”); see also Gibson v. Cendyn Grp., LLC, No. 23-16463 (9th Cir. 2025) (distinguishing software that shares confidential information among competitors from software that does not).
[43] Roger Alford, Deputy Assistant Att’y Gen., U.S. Dep’t of Justice, The Role of Antitrust in Promoting Innovation 8 (Feb. 23, 2018).
[44] Brief of the International Center for Law & Economics as Amicus Curiae, supra note 42, at 6.
[45] William E. Kovacic et al., Plus Factors and Agreement in Antitrust Law, 110 Mich. L. Rev. 393 (2011); 6 Phillip E. Areeda & Herbert Hovenkamp, Antitrust Law: An Analysis of Antitrust Principles and Their Application ¶ 1433 (5th ed. 2020).
[46] Recent litigation involving platform-pricing policies includes Gibson v. Cendyn Grp., LLC, No. 23-16463 (9th Cir. 2025); United States v. RealPage, Inc., No. 1:24-cv-00710 (M.D.N.C.); and FTC v. Amazon.com, Inc., No. 2:23-cv-01495 (W.D. Wash.).
[47] The economic literature identifies several concrete procompetitive mechanisms, including free-rider prevention and investment protection. See Benjamin Klein, Competitive Resale Price Maintenance in the Absence of Free Riding, 76 Antitrust L.J. 431 (2009); Chengsi Wang & Julian Wright, Search Platforms: Showrooming and Price Parity Clauses, 51 Rand J. Econ. 32 (2020). For the principal academic argument against MFNs, see Jonathan B. Baker & Fiona Scott Morton, Antitrust Enforcement Against Platform MFNs, 127 Yale L.J. 2176 (2018). The empirical literature, however, has not borne out the harms they predicted. See, e.g., infra note 49.
[48] Blue Cross & Blue Shield of Wis. v. Marshfield Clinic, 65 F.3d 1406, 1415 (7th Cir. 1995) (Posner, J.); see also Kartell v. Blue Shield of Mass., 749 F.2d 922 (1st Cir. 1984) (Breyer, J.).
[49] See, e.g., Andrea Mantovani, Claudio A. Piga & Carlo Reggiani, Online Platform Price Parity Clauses: Evidence from the EU Booking.com Case, 69 J. Indus. Econ. 422 (2021) (finding only an approximately 2.6% short-run price decrease from removing MFNs, with no statistically significant medium-run effects); Jack (Peiyao) Ma et al., The Price Effects of Prohibiting Price Parity Clauses: Evidence from Global Hotel Chains, Econ. J., (forthcoming) (finding no significant price effects on visible online channels after prohibiting all price-parity clauses, with consumer-welfare gains attributable primarily to channel shifting rather than competitive pricing effects).
[50] United States v. Apple, Inc., 791 F.3d 290 (2d Cir. 2015), cert. denied, 136 S. Ct. 1158 (2016).
[51] Id.
[52] Benjamin Klein, The Apple E-Books Case: When Is a Vertical Contract a Hub in a Hub-and-Spoke Conspiracy?, 13 J. Competition L. & Econ. 423 (2017).
[53] Howard Hess Dental Labs., Inc. v. Dentsply Int’l, Inc., 602 F.3d 237, 244 (3d Cir. 2010) (“The rim of the wheel is the connecting agreements among the horizontal competitors … that form the spokes.”).
[54] Bell Atl. Corp. v. Twombly, 550 U.S. 544, 553–54 (2007) (parallel conduct “just as much in line with a wide swath of rational and competitive business strategy unilaterally prompted by common perceptions of the market”).
[55] See Kevin R. Williams, The Welfare Effects of Dynamic Pricing: Evidence from Airline Markets, 90 Econometrica 831 (2022); Juan Camilo Castillo, Who Benefits from Surge Pricing?, 93 Econometrica 1811 (2025).
[56] Org. for Econ. Cooperation & Dev. (OECD), Algorithmic Competition 14 (2023), https://www.oecd.org/daf/competition/algorithmic-competition2023.pdf.
[57] Cody Taylor, The Case for Algorithmic Pricing: Consumer Welfare, Market Efficiency, and Policy Missteps, Mercatus Ctr. Pol’y Brief (May 14, 2025), https://www.mercatus.org/research/policy-briefs/case-algorithmic-pricing-consumer-welfare-market-efficiency-and-policy.
[58] See Cont’l T.V., Inc. v. GTE Sylvania Inc., 433 U.S. 36 (1977); Leegin Creative Leather Prods., Inc. v. PSKS, Inc., 551 U.S. 877 (2007); Lester G. Telser, Why Should Manufacturers Want Fair Trade?, 3 J.L. & Econ. 86 (1960).
[59] Stephanie Assad, Robert Clark, Daniel Ershov & Lei Xu, Algorithmic Pricing and Competition: Empirical Evidence from the German Retail Gasoline Market, CESifo Working Paper No. 8521 (2020).
[60] See, e.g., Husanjot Chahal, Helen Toner & Ilya Rahkovsky, Small Data’s Big AI Potential, Ctr. for Sec. & Emerging Tech. (2021), https://cset.georgetown.edu/wp-content/uploads/CSET-Small-Datas-Big-AI-Potential-1.pdf; Josh Howard & Amit Kara, Data Silos Explained: Problems They Cause and Solutions, Databricks (Nov. 11, 2024), https://www.databricks.com/blog/data-silos-explained-problems-they-cause-and-solutions.
[61] Alden Abbott & Satya Marar, Is Data Really a Barrier to Entry? Rethinking Competition Regulation in Generative AI, Mercatus Ctr. Working Paper (Mar. 31, 2025), https://www.mercatus.org/research/working-papers/data-really-barrier-entry-rethinking-competition-regulation-generative-ai.
[62] European Comm’n, Guidelines on the Applicability of Article 101 of the Treaty on the Functioning of the European Union to Horizontal Co-operation Agreements, 2023 O.J. (C 259) 1, ¶¶ 390–450 [hereinafter 2023 EU Horizontal Guidelines].
[63] U.S. Dep’t of Justice & Fed. Trade Comm’n, Antitrust Policy Statement on Sharing of Cybersecurity Information (2014), https://www.ftc.gov/system/files/documents/public_statements/297681/140410ftcdojcyberthreatstmt.pdf.
[64] Cybersecurity Information Sharing Act of 2015, Pub. L. No. 114-113, div. N, tit. I, 129 Stat. 2935, 2947–59 (2015).
[65] Fed. Trade Comm’n & U.S. Dep’t of Justice, Statements of Antitrust Enforcement Policy in Health Care (Aug. 1996), https://www.ftc.gov/system/files/attachments/competition-policy-guidance/statements_of_antitrust_enforcement_policy_in_health_care_august_1996.pdf; see also Fed. Trade Comm’n, Press Release, Federal Trade Commission Withdraws Health Care Enforcement Policy Statements (July 14, 2023), https://www.ftc.gov/news-events/news/press-releases/2023/07/federal-trade-commission-withdraws-health-care-enforcement-policy-statements.
[66] 2023 EU Horizontal Guidelines, supra note 62, ¶¶ 552–584.
[67] U.S. Dep’t of Justice & Fed. Trade Comm’n, Antitrust Guidelines for Business Activities Affecting Workers (Jan. 2025), https://www.ftc.gov/system/files/ftc_gov/pdf/p251201antitrustguidelinesbusinessactivitiesaffectingworkers2025.pdf.
[68] Aya Healthcare Servs., Inc. v. AMN Healthcare, Inc., 9 F.4th 1102 (9th Cir. 2021).
[69] NCAA v. Alston, supra note 15.
[70] United States v. Am. Airlines Grp. Inc., 121 F.4th 209 (1st Cir. 2024).
[71] See Petition for Writ of Certiorari, Am. Airlines Grp. Inc. v. United States, No. 24-938 (U.S. Feb. 27, 2025), https://www.supremecourt.gov/DocketPDF/24/24-938/350862/20250227152404466_2027-02-27%20American%20Airlines%20Petition%20For%20Certiorari%20with%20Appendix.pdf.
[72] Am. Airlines Grp. Inc. v. United States, No. 24-938 (U.S. June 30, 2025) (mem.).
[73] Commission Regulation 2023/1066, supra note 20; Commission Regulation 2023/1067, supra note 22.
[74] 2023 EU Horizontal Guidelines, supra note 62.
[75] Id. at 18–19, 24–25 (discussing grace periods).