global geopolitics

BlackRock, Vanguard and State Street sit on both sides of the copyright bargain, which makes voluntary consent an accounting entry rather than a settlement with creators.

Editorial Analysis | October 2026


The New York Times Company filed suit against Microsoft and OpenAI in December 2023 alleging that its journalism had been reproduced without licence or payment. Vanguard and BlackRock rank among the largest institutional holders of the plaintiff, and both rank among the largest holders of Microsoft and of the venture-backed developer whose investors include the same asset managers. The litigation is therefore a dispute between two firms whose residual owners substantially overlap, which reframes the question of consent as a question of internal accounting. Observers who treat the case as evidence that markets are correcting the training-data problem have looked at the parties in the courtroom without looking at the shareholders behind them. Corporate filings disclose that arrangement plainly, which makes the omission a matter of analytical habit rather than evidentiary scarcity.

Common ownership research supplies the framework that most commentary on AI governance conspicuously avoids. Fichtner, Heemskerk and Garcia-Bernardo documented the re-concentration of corporate ownership through passive index funds and identified the Big Three asset managers as pervasive blockholders across competing firms (Fichtner, Heemskerk and Garcia-Bernardo, 2017). Azar, Schmalz and Tecu subsequently demonstrated that horizontal shareholding within the airline industry produced pricing effects inconsistent with independent rivalry (Azar, Schmalz and Tecu, 2018). The mechanism operates through diversified portfolios rather than collusive intent, because an owner holding both sides of a transaction is indifferent to which side captures the surplus. Licensing payments between a developer and a publisher therefore represent a transfer from one part of a portfolio to another part of the same portfolio, minus transaction costs and advisory fees.

Individual creators occupy a structurally different position because they own neither side of that transaction and possess no bargaining unit capable of matching corporate counterparties. The corpus that trains contemporary models consists overwhelmingly of work produced by journalists, photographers, novelists, translators, musicians, and software engineers who assigned their rights through employment contracts and platform terms. Licensing deals negotiated with large publishers and image libraries compensate the entities that aggregated those rights rather than the people who generated the underlying material. Collective management through statutory remuneration, extended collective licensing, and data trusts offers the only mechanisms capable of reaching individual authors at scale, and none of those mechanisms appears in the voluntary agreements currently celebrated as market solutions (Directive (EU) 2019/790; Intellectual Property Office, 2022). Consent without a bargaining agent is a legal formality rather than an economic settlement.

The compute layer reinforces the same configuration because the capital that finances data centres is substantially the capital that holds the developers and the publishers. BlackRock, Global Infrastructure Partners, Microsoft and MGX announced an AI infrastructure partnership in September 2024 targeting thirty billion dollars of equity investment and up to one hundred billion dollars including debt (BlackRock, 2024). BlackRock had completed its acquisition of Global Infrastructure Partners earlier that year, which placed a single asset manager at the centre of data centre, pipeline, and power generation ownership. Electricity supply agreements, water rights, and land acquisition for hyperscale facilities now proceed through the same institutional holders whose portfolios contain the media companies negotiating training licences. The means of computation and the means of cultural production have converged into common beneficial ownership while the policy conversation treats them as separate industries.

Lobbying expenditure and personnel circulation provide the second structural channel through which that ownership translates into regulatory outcomes. Corporate Europe Observatory documented the AI Act as the most heavily lobbied file in the European Union’s history, with developer representatives meeting Commission officials and parliamentarians at a volume exceeding most financial services files (Corporate Europe Observatory, 2023). American disclosure filings show sustained increases in federal lobbying by the largest developers and their trade associations across the same period (OpenSecrets, 2025). Revolving-door appointments move officials between the developers, the standards bodies, the national AI institutes, and the offices that draft procurement rules. Michael Kratsios moved from the White House technology office to a senior role at Scale AI and subsequently returned to direct science and technology policy, which illustrates a trajectory that has become routine rather than exceptional (White House, 2025).

The think tank and political finance layer completes the circuit because the same firms fund the intellectual infrastructure within which legislative options are formulated. Technology firms have financed research institutions, scholarship programmes, and policy centres whose output defines the range of respectable positions on AI regulation, and the withdrawal of that funding has produced documented ruptures when recipients departed from donor preferences. Trade associations aggregate corporate political spending and channel it through vehicles that obscure individual contributions while advancing a common legislative agenda. The World Economic Forum’s AI Governance Alliance convenes developers, asset managers, and regulators within a framework that presupposes multistakeholder consensus rather than democratic contestation (World Economic Forum, 2023). Multistakeholderism functions as a governance doctrine that distributes authority according to convening power rather than according to electoral mandate.

The judiciary has independently narrowed the space within which regulatory agencies might act, and that development deserves more weight than it commonly receives in AI policy commentary. The United States Supreme Court’s decision in Loper Bright Enterprises against Raimondo overruled the Chevron deference doctrine and transferred interpretive authority from agencies to courts (Loper Bright Enterprises v. Raimondo, 2024). Litigants challenging AI rules can now forum-shop toward circuits whose judges hold settled preferences against administrative authority, which raises the expected cost of enforcement above the level legislatures intended. Arbitration clauses and platform terms of service further remove disputes from public adjudication and place them before forums selected by the defendant. The courts have therefore become an active participant in regulatory design rather than a neutral arbiter of legislative intent.

Regulatory capture theory describes these channels with considerable precision, and the AI sector exhibits several of them simultaneously rather than sequentially. Stigler and Peltzman established that concentrated industries with strong per-firm stakes will outorganise diffuse publics in the competition to shape rules (Stigler, 1971; Peltzman, 1976). Carpenter and Moss catalogued the mechanisms through which capture operates, including funding dependence, information asymmetry, cultural convergence, and revolving doors (Carpenter and Moss, 2014). The European Union’s reliance on harmonised standards drafted with heavy industry participation converts technical committees into sites where market structure is determined under the appearance of engineering neutrality (Regulation (EU) 2024/1689). Incumbents have every incentive to shape standards that raise rivals’ compliance costs, and the delay in delivering those standards has already advantaged firms with internal compliance capacity.

Voluntary commitments operate within this configuration rather than against it, and their principal function is to occupy the political space that binding rules might otherwise fill. The Bletchley Declaration, the Seoul Frontier AI Safety Commitments, and the G7 Hiroshima Process code of conduct establish norms without verification, penalties, or independent audit authority (AI Safety Summit, 2023; AI Seoul Summit, 2024; G7, 2023). Firms publish capability and safety information selectively, on timelines they control, using methodologies they commission. The result is a disclosure regime in which the regulated party determines what counts as a relevant fact, which is a definitional feature of capture rather than an unfortunate lapse. Treating such commitments as evidence of good faith requires ignoring who writes them and who benefits from their non-binding character.

Independent oversight commissions can be designed to resist these pressures, and the design features that matter are well established in the comparative literature on regulatory institutions. Levy-based funding through assessed contributions on compute, revenue, or data processing removes dependence on annual appropriations and on the governments that champion national champions. Fixed statutory terms, published appointment criteria, and cooling-off periods constrain the revolving door that otherwise imports industry preferences into supervisory judgement. Statutory access to training data documentation, model weights for audit purposes, incident reports, and internal evaluation records converts transparency from a voluntary gesture into an enforceable obligation. Complaint procedures with standing for affected workers, artists, and communities, together with whistleblower protections and fee-shifting provisions, give the public practical routes for contesting decisions. None of these mechanisms requires technical knowledge that regulators cannot acquire, and all of them require political power that legislatures have so far declined to exercise.

The deeper question concerns ownership rather than procedure, because supervisory bodies operate within an economy whose commanding heights have consolidated into a small number of asset managers and platform firms. Antitrust remedies addressed to horizontal shareholding, structural separation of computation from content, public compute facilities offered on non-discriminatory terms, and public data trusts holding licensed corpora under collective governance each attack the configuration rather than its symptoms. A statutory remuneration right for authors whose work enters training corpora would create a revenue stream independent of publisher bargaining power. Procurement conditionality through the state’s own purchasing power can impose auditability requirements that legislation struggles to enact. These instruments exist in other sectors and require only the political will that currently attaches to industrial policy rather than to oversight.

The trajectory most consistent with present evidence involves continued co-regulation in Europe, deregulatory oscillation in the United States, and standards harmonisation that consolidates incumbent advantage while producing modest transparency gains. Litigation will generate settlements and licensing fees that flow to aggregators rather than to creators, and those settlements will be cited as proof that markets have resolved the consent problem. Personnel will continue to circulate between developers, regulators, and international secretariats, and each rotation will be described as an exercise in expertise rather than as a transfer of influence. Whether independent oversight becomes a meaningful constraint or a ceremonial supplement depends on whether legislatures construct funding, access, and appointment rules capable of surviving the political pressure that the firms concerned can mobilise. The answer will be determined by class organisation rather than by regulatory technique, because the ownership of computation and of cultural production now rests in the same hands.



Authored By: Global GeoPolitics

Thank you for visiting. This is a reader-supported publication. If you believe journalism should serve the public, not the powerful, and you’re in a position to help, becoming a PAID SUBSCRIBER truly makes a difference. Alternatively you can support by way of a cup of coffee:

https://buymeacoffee.com/ggtv |

https://ko-fi.com/globalgeopolitics |

Bitcoin: 3NiK8BoRZnkwJSHZSekuXKFizGPopkE7ns



Reference List

AI Safety Summit (2023) The Bletchley Declaration by Countries Attending the AI Safety Summit, 1–2 November 2023. London: Department for Science, Innovation and Technology.

AI Seoul Summit (2024) Frontier AI Safety Commitments, 21 May 2024. Seoul: Ministry of Science and ICT.

Azar, J., Schmalz, M. C. and Tecu, I. (2018) ‘Anticompetitive effects of common ownership’, Journal of Finance, 73(4), pp. 1513–1565.

BlackRock (2024) BlackRock, Global Infrastructure Partners, Microsoft and MGX launch new AI infrastructure investment partnership. Press release, 17 September. New York: BlackRock.

Carpenter, D. and Moss, D. A. (eds.) (2014) Preventing Regulatory Capture: Special Interest Influence and How to Limit It. Cambridge: Cambridge University Press.

Corporate Europe Observatory (2023) Big Tech’s lobbying frenzy on the AI Act. Brussels: Corporate Europe Observatory.

Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on copyright and related rights in the Digital Single Market, OJ L 130, 17.5.2019, pp. 92–125.

Fichtner, J., Heemskerk, E. M. and Garcia-Bernardo, J. (2017) ‘Hidden power of the Big Three? Passive index funds, re-concentration of corporate ownership, and new financial risk’, Business and Politics, 19(2), pp. 298–326.

G7 (2023) Hiroshima Process International Code of Conduct for Organizations Developing Advanced AI Systems. Tokyo: G7 Presidency.

Intellectual Property Office (2022) Artificial Intelligence and Intellectual Property: Consultation on Copyright and Related Rights. London: Intellectual Property Office.

Loper Bright Enterprises v. Raimondo, 603 U.S. 369 (2024).

Microsoft Corporation (2024) Definitive Proxy Statement (Schedule 14A). Washington, DC: Securities and Exchange Commission.

OpenSecrets (2025) Lobbying profile: OpenAI. Washington, DC: Center for Responsive Politics.

Peltzman, S. (1976) ‘Toward a more general theory of regulation’, Journal of Law and Economics, 19(2), pp. 211–240.

Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence, OJ L, 12.7.2024.

Stigler, G. J. (1971) ‘The theory of economic regulation’, Bell Journal of Economics and Management Science, 2(1), pp. 3–21.

The New York Times Company v. Microsoft Corporation, OpenAI Inc. et al., No. 1:23-cv-11195 (S.D.N.Y. filed 27 December 2023).

The New York Times Company (2024) Definitive Proxy Statement (Schedule 14A). Washington, DC: Securities and Exchange Commission.

White House (2025) President Trump announces appointment of Michael Kratsios as Director of the Office of Science and Technology Policy. Washington, DC: The White House.

World Economic Forum (2023) AI Governance Alliance. Geneva: World Economic Forum.