Global Policy · en · 10 min

By Caroline V. Beaumont · April 10, 2026

As AI-driven optimization reshapes how assets are allocated, prices discovered, and information flows, the public markets stand at a pivotal juncture for a…

As AI-driven optimization reshapes how assets are allocated, prices discovered, and information flows, the public markets stand at a pivotal juncture for antitrust and competition policy. This piece examines how AI-enabled decision engines influence market power, the regulatory responses underway, and the risks and opportunities for fairness, efficiency, and resilience in global capital markets.

AI-augmented trading and concentration of market power

The deployment of machine learning in trading platforms, execution venues, and clearinghouses accelerates price discovery and liquidity provision while potentially concentrating power among a small cohort of sophisticated players. As of late 2025, industry analyses show that high-frequency trading firms account for roughly 50–70% of U.S. equity order flow on the busiest exchanges during peak hours, with similar patterns in several major European markets. In the same window, AI-driven routing optimizes cross-venue execution, and some studies report that AI-optimized orders can reduce effective spreads by between 12–22% on liquid mid-cap stocks, but with the caveat that a small number of AI-enabled optimizers control a disproportionate share of steady-state liquidity. The risk is not merely price discipline; it is the potential for tacit collusion and reduced entry incentives for new trading venues that cannot match the AI-enabled efficiency of incumbents.

Regulators are grappling with the way algorithms create feedback loops. In the 2024 EU AI Act, emphasis was placed on governance around decision-making systems that influence market outcomes, including model risk management, auditability, and the potential for regime-driven herding. The Financial Stability Board, in its 2025 assessment, warned that systemic liquidity fragility could rise if AI-driven liquidity providers simultaneously withdraw during stress periods, amplifying flash crashes. U.S. market regulators have proposed enhanced pre-trade risk controls for automated traders, and several central banks are piloting SAMPs (synthetic agent market simulations) to stress-test AI-driven strategies under extreme conditions. The core concern is that AI can compress the time horizon for strategic decision-making to milliseconds, eroding the ability of human-in-the-loop oversight to intervene before cascading effects unfold.

Vertical integration of data and the antitrust question

AI systems rely on vast data inputs, and the data footprint of large financial institutions is a critical determinant of predictive accuracy and leading-edge optimization. As of 2024–2025, the top five global banks and top three fintech platforms collectively controlled an estimated 70–85% share of high-fidelity market data in several major markets, with customer trade data and permissioned datasets driving a competitive edge for model training. Firms that combine market data, proprietary analytics, and execution capabilities can extract rents not just from price discovery but from forecasting accuracy that determines order placement and liquidity provision.

Antitrust scrutiny has focused on potential monopolistic leverage through data hoarding and gatekeeping. Regulators in the EU are examining whether data-sharing obligations among market participants should be mandated to prevent a "data moat" from locking in AI advantages. In the United States, several proceedings contemplated under the 2023–2025 climate of antitrust enforcement include whether large financial-services platforms’ control of data ecosystems constitutes an unlawful tying of services or an unlawful combination that forecloses rivals. Competition authorities increasingly articulate data access as a core competition lever, though the operationalizing of this principle across global markets remains uneven due to differences in data privacy regimes, sectoral exemptions, and concerns about operational risk from mandated data provision.

Regulatory tools: from disclosure to ex-ante controls

The policy toolkit for AI in public markets spans disclosure, governance, risk management, and ex-ante constraints. One trend is the expansion of model governance frameworks that require firms to document training data provenance, model architecture, and update cadence. By late 2025, several jurisdictions introduced or deepened requirements for AI systems used in critical market functions to publish annual model-impact reports, including the scope of data sources, error rates, and governance checks. For example, a cross-border framework in a 2025 compact requires semi-annual public disclosure of model performance metrics relative to a standardized benchmark and quarterly disclosures of any material model drift or retraining events affecting risk controls. In practice, these measures aim to reduce information asymmetries that can obscure the true sources of competition or enable predatory pricing through opaque optimization strategies.

• Ex-ante transaction-controls: Several regulators have proposed or implemented limits on the rate of automated orders or on the maximum size of AI-augmented execution blocks during periods of volatility to curb rapid feedback loops.
• Mandatory data sharing: Proposals to standardize data schemas and enable access to non-proprietary datasets are advancing, with emphasis on fairness, non-discrimination in access, and privacy safeguards.
• Auditability and explainability: Institutions face increasing expectations to demonstrate that AI decision processes either produce demonstrably fair outcomes or that any harmful biases are mitigated with robust controls.

However, ex-ante controls risk dampening genuine efficiency gains. If safeguards are overly burdensome or misaligned with the pace of market innovation, there is a danger of driving competitive activity into jurisdictions with lighter frameworks, raising regulatory arbitrage concerns. The challenge is to calibrate rules so they promote resilience and fairness without throttling beneficial optimization. Early empirical signals suggest that jurisdictions adopting proportionate disclosure accompanied by real-time risk monitoring have seen faster adoption of responsible AI practices, though the impact on market concentration remains mixed depending on the structure of each market and the stringency of compliance regimes.

Vertical integration vs. interoperability: consequences for entry and innovation

The AI-enabled optimization stack tends to favor players with end-to-end capabilities—from data ingestion and model development to execution and settlement. This vertical integration can suppress entry by new platforms that excel in one layer but cannot compete across the entire stack. As of 2025, surveys of active market participants indicate that roughly 60–70% of AI-driven liquidity provision is concentrated among a handful of vertically integrated intermediaries in both North American and European markets. While some observers celebrate the efficiency gains of such integration, antitrust advocates warn that deep vertical ties can reduce the incentive for rivals to invest in complementary capabilities, thereby raising switching costs and reducing dynamic competition.

Interoperability initiatives carry potential benefits. Open APIs for order routing, standard risk controls, and shared reference data can lower the barriers to entry for new entrants and smaller players, enabling them to compete on service quality and pricing rather than sheer data volume. In the 2024 EU Digital Markets Act and the 2025 amendments to the US Market Access Rules, policymakers signaled willingness to encourage interoperability where it enhances contestability. Yet interoperability must be designed with safeguards to prevent information leakage that could enable unfair advantages, such as models trained on aggregations of rival strategies or sensitive client-private data. The balance is to foster a modular ecosystem that preserves incentives for innovation while preventing a data or execution monopoly from crystallizing around a single consortium of providers.

Quantitative signals in late 2025 show that markets with higher interoperability scores report more entrants into high-speed trading segments and a broader array of venue choices, with some exchanges reporting a 12–18% increase in new liquidity providers after interoperability pilot programs. Conversely, markets slow to adopt open interfaces experienced slower diversification and higher execution costs for active traders, suggesting that the governance of interoperability is as important as the technical design. The risk of “platform lock-in” remains real, particularly for institutions that rely on bespoke tools built on tightly coupled AI stacks.

Global policy harmonization: lessons from diverse regulatory ecosystems

Markets are global, yet regulatory approaches diverge in important ways. The EU’s 2024–2025 regime emphasizes transparency, data governance, and accountability for algorithmic decision-making in financial services, while the United States emphasizes sector-specific risk controls, enforcement discretion, and market-structure oversight. The Asia-Pacific landscape shows a mix of rigorous data-protection regimes and rapid experimentation in AI governance, with Singapore, Hong Kong, and Australia pursuing risk-based licensing for AI-enabled trading and data services. As of late 2025, policymakers increasingly acknowledge that cross-border cooperation is essential to curb regulatory arbitrage and to address the transnational effects of AI-driven market power. Initiatives include coordinated model risk management standards, shared incident reporting on AI-driven market anomalies, and joint exercises to stress-test AI-enabled liquidity providers under synchronized shocks.

One practical implication is the need for consistent disclosure standards that transcend borders. If a firm operates in multiple jurisdictions, disparate reporting requirements can create uneven playing fields. A recent cross-jurisdictional synthesis suggests that a unified baseline for model governance—covering data provenance, update cadence, performance disclosure, and risk controls—could reduce compliance fragmentation and improve comparability for policymakers. The challenge remains aligning privacy protections with the need for data provenance and access to non-proprietary datasets. The 2025 NFPA 1500 update, which touches on human factors in fire-and-emergency services but is cited here as an example of how cross-disciplinary standards are evolving, underscores a broader trend: regulatory architectures benefit from modular, auditable components that can be adapted to rapid AI advances without sacrificing safety, privacy, or competition goals.

Global policy momentum matters for resilience. When AI-driven trading systems are vulnerable to coordinated failures or external manipulation, the systemic risk footprint crosses borders quickly. The 2025 Central Bank–Regulator Forum produced a consensus that cross-border incident reporting and mutual assistance arrangements should be improved, with a particular emphasis on AI-driven market-making networks. The practical takeaway is clear: policy design must prioritize cross-border data flows with appropriate safeguards, establish shared testing environments, and maintain the ability to apply proportionate remedies that preserve market functioning while minimizing harm to competition and consumer protection.

Consumer and market fairness: spillovers beyond institutional players

AI in public markets does not only affect institutions; it changes the experience of everyday investors and the integrity of price formation. Transparency deficits around execution quality and model-driven order routing can undermine trust in markets, even when headline metrics appear efficient. As of 2025, several jurisdictions have mandated enhanced disclosure of execution metrics, including latency, slippage, and fill rates, to help investors gauge whether AI-enabled optimization benefits are shared with end users or captured by a narrow professional class. Preliminary data from major market centers show that retail participation remains robust but is unevenly distributed: in some markets, households using high-speed trading platforms experience lower spreads on liquid assets but face higher costs in less liquid segments, exacerbating inequality in access to efficient execution.

• Execution quality: In high-liquidity equities, average effective spreads for AI-optimized orders narrowed by approximately 8–15% compared with conventional orders in 2024–2025, yet in mid-cap and small-cap segments the disparity persisted, with AI-enabled paths occasionally widening latency-induced costs during volatile periods.
• Transparency: Early 2025 surveys indicate that 60–70% of retail traders reported limited visibility into which venues benefited most from AI routing, highlighting a need for clearer disclosure regimes and independent verification of exchange performance claims.

Policy responses must balance efficiency gains with protections against predatory practices. Stronger provisions around best-execution obligations, independent verification of routing logic, and caps on potential conflicts of interest in data provisioning can help ensure that AI-driven improvements translate into tangible benefits for a broad investor base—not just market insiders. In doing so, regulators support the integrity of price discovery, which ultimately underpins trust in the capital markets and the broader economy.

Conclusion

AI-enabled optimization promises meaningful gains in efficiency, liquidity, and price discovery, but it also reshapes market power in ways that challenge traditional antitrust tools. The central tasks for policymakers are to prevent entrenched advantages born of data and control over AI-enabled decision stacks, to foster interoperability that preserves contestability, and to align cross-border regulatory regimes in a manner that reduces coordination failures and regulatory arbitrage. As of late 2025, a measured combination of governance, disclosure, and targeted ex-ante controls—balanced with spaces for legitimate competitive experimentation—appears best suited to harness AI’s potential while guarding against dampened competition, systemic risk, and inequitable access to efficient markets. The stakes are high: a robust, competitive public market ecosystem depends on policies that anticipate the speed of AI innovation without sacrificing accountability, transparency, or fairness.