Conditionality (Mazzucato)
Conditionality is central to Mazzucato's institutional vocabulary for the AI transition. It means that public support — tax advantages, procurement contracts, regulatory accommodation, access to publicly funded research — should come with requirements attached: requirements for data sharing, interoperability, fair pricing, environmental disclosure, worker protections, and distribution of gains. The current arrangement, in which public support flows to AI companies unconditionally, represents the absence of institutional design rather than a principled choice. Applied to the AI economy, conditionality would redirect innovation — ensuring that technologies developed with public support, trained on public data, and deployed on public infrastructure serve public purposes as well as private ones. Mazzucato has deployed the concept with particular force in her critique of the UK AI Action Plan, arguing that governments must move beyond unbalanced relationships with digital monopolies and stop offering technology companies lucrative unstructured deals with no conditionalities attached.
The Coordination Ceiling Problem — Contrarian ^ Opus
There is a parallel reading that begins from the institutional capacity required to design, enforce, and update conditionality frameworks at the pace AI systems evolve. The precedents Mazzucato cites—defense procurement, pharmaceutical pricing, renewable energy mandates—operate in domains where the technical substrate changes slowly enough that regulators can develop expertise, where the relevant metrics are measurable (drug efficacy, domestic content percentages, emissions reductions), and where the relevant timescales match bureaucratic cycles. AI capability develops on six-month horizons; regulatory capacity develops on six-year ones. The expertise asymmetry is profound: the firms understand their systems in ways regulators cannot match, the interoperability requirements that sound reasonable in principle encounter genuine technical obstacles the public sector lacks the capacity to adjudicate, and the enforcement mechanisms require sustained political will that outlasts electoral cycles.
The political economy cuts the other direction as well. Conditionality frameworks create new surfaces for capture—the interoperability standards bodies become venues for incumbent advantage, the transparency requirements generate compliance burdens that favor large firms over challengers, the participatory governance mechanisms get dominated by the most organized builder constituencies rather than the most affected populations. The Nordic examples Mazzucato invokes operated in economies with strong labor movements, high social trust, and limited exit options for capital; the AI economy features footloose global firms, weak worker organization in the relevant sectors, and jurisdictions competing to offer the most favorable terms. Conditionality in this context risks becoming theater: symbolic requirements that create the appearance of public direction while the substantive decisions remain private.
In the AI Story
The application of conditionality to AI platforms would produce specific institutional mechanisms. Mandatory interoperability requirements ensure that builders can move between platforms without losing their work, data, or workflows. Data portability standards enable builders to extract their contributions from one platform and import them to another. Transparency requirements obligate platforms to disclose criteria for pricing decisions, feature availability, terms of service changes, and algorithmic ranking. Participatory governance mechanisms — builder advisory councils with formal standing, mandatory consultation before significant changes, independent arbitration for disputes — give builders a structured voice.
Conditionality has substantial precedent in public procurement and industrial policy. Defense contractors face extensive conditions on their procurement contracts. Pharmaceutical companies receiving public research grants face conditions on pricing of resulting products (albeit inadequately enforced). Renewable energy subsidies in many jurisdictions come with domestic content and labor requirements. The principle that public money carries public obligations is well-established; its application to the AI economy is a matter of extending existing precedent rather than inventing novel mechanisms.
The shift to conditionality is fundamental rather than incremental. The current arrangement treats AI companies as recipients of public support whose obligations to the public are minimal — a tax contribution proportional to their (often minimized) profits, compliance with applicable regulations. Conditionality reframes the relationship: the public provides the foundational research, training data, infrastructure, and workforce that make AI capability possible, and these contributions generate legitimate claims on the behavior of firms that commercialize the capability.
The political economy of implementing conditionality is challenging. AI companies have strong incentives to resist requirements that constrain their operational flexibility or reduce their capacity to capture surplus. The counterargument — that conditionality reduces innovation incentives — has been deployed against every conditionality framework in the history of industrial policy and has been empirically refuted by the superior innovation performance of economies with more extensive conditionality frameworks (the Nordic countries, Germany, South Korea).
Origin
Conditionality as a policy principle has roots in classical industrial policy — Hamilton's 1791 Report on Manufactures, the infant industry protection tradition, the Nordic developmental state. Mazzucato's contribution is its systematic application to innovation policy and its development as a framework for the platform and AI economies.
Her most direct applications to AI have come in her critique of the UK AI Action Plan (2024), the Project Syndicate essays with Tommaso Valletti (2025), and the ongoing Algorithmic Rents research program.
Key Ideas
Public money, public obligations. Support flows come with enforceable requirements — the principle is well-established in industrial policy.
Operational mechanisms. Interoperability, portability, transparency, participatory governance — specific requirements rather than vague commitments.
Precedent exists. Defense procurement, pharmaceutical grants, renewable energy — conditionality has been applied in multiple domains without destroying the industries affected.
Redirection of innovation. Conditionality does not reduce innovation; it redirects innovation toward public purposes.
Current absence is design gap. The unconditional support for AI companies is an institutional omission, not a principled choice.
Debates & Critiques
Critics argue conditionality would slow AI deployment during a critical period of international competition. Mazzucato's response is that conditionality is precisely what enables sustainable deployment by ensuring the political legitimacy and distributional adequacy that unconditional deployment cannot secure.
Appears in the Orange Pill Cycle
Domain-Specific Feasibility Gradients — Arbitrator ^ Opus
The right weighting depends entirely on which dimension of conditionality you're examining. On the question of whether public support should generate public obligations in principle—Mazzucato's framework is fully correct (100%). The asymmetry where public research, infrastructure, and data flow to private firms that owe minimal obligations back is indeed a design gap rather than a necessity. On whether specific conditionality mechanisms (interoperability mandates, data portability standards) are technically feasible—the answer varies by domain. Interoperability for model APIs is achievable (70% Mazzucato); interoperability for training pipelines faces genuine coordination problems (40% Mazzucato, 60% contrarian). The expertise asymmetry is real but not insurmountable—public sector technical capacity can be built, though the contrarian is right (70%) that current institutional structures are inadequate to the pace and complexity.
On the political economy question—both views hold partial truth in ways that suggest a different frame. The capture risk the contrarian names is real (they're right at 65%), but it points toward the design of conditionality frameworks rather than their rejection: transparent standard-setting processes, automatic sunset provisions, built-in review cycles. Mazzucato is right (75%) that the comparative performance of economies with stronger conditionality shows the framework doesn't destroy innovation, but the contrarian is right (60%) that those examples operated under different structural conditions—higher trust, stronger labor, less capital mobility.
The synthetic frame: conditionality as a portfolio of mechanisms rather than a single intervention. Some dimensions (pricing transparency, procurement requirements) are highly feasible now; others (algorithmic accountability, participatory governance) require institution-building first. The question isn't conditionality yes/no—it's which conditions, enforced how, with what adjustment mechanisms.
Further reading
- Mazzucato, Mariana. Critique of the UK AI Action Plan. IIPP Working Paper, 2024.
- Chang, Ha-Joon. Kicking Away the Ladder. Anthem Press, 2002.
- Wade, Robert. Governing the Market. Princeton, 1990.